개발자로서 현장에서 일하면서 새로 접하는 기술들이나 알게된 정보 등을 정리하기 위한 블로그입니다. 운 좋게 미국에서 큰 회사들의 프로젝트에서 컬설턴트로 일하고 있어서 새로운 기술들을 접할 기회가 많이 있습니다. 미국의 IT 프로젝트에서 사용되는 툴들에 대해 많은 분들과 정보를 공유하고 싶습니다.
※ 본 리그는 메가존 클라우드가 자체적으로 진행하는 행사로 AWS에서 개최하는 리그와 무관합니다. ※ 상금 및 경품 지급 시 소득세 등 제세공과금이 차감 혹은 청구 됩니다. ※ 본 경기 규칙은 AWS DeepRacer League 규칙을 따르며 트랙 또한 re:Invent 2018 트랙에서 진행 됩니다.
========================================
저는 AWS Deepracer 모델 차량을 7월 중순에 받을 예정이라서 10월 3일 토론토에서 열리는 경기에 참가할 수 있을 것 같습니다.
휴가 내고 비행기 타고 가서 참가할 생각인데....
어떻게 될 지 아직......
지금 제가 만들고 있는 모델은 Kumo Torakku 트랙에서 23초를 기록하고 그 이후에는 전혀 기록이 나아 지질 않고 있습니다.
지금 1,2,3 등은 모두 10초 대 이던데.... 그런 기록은 어떻게 하면 낼 수 있을 지.......
re:Invent 2019에서 AWS DeepRacer Cup에서 우승해 상금, 영예, AWS DeepRacer Championship Cup을 획득하세요. 매월 열리는 가상 Circuit 레이스에서 온라인으로 경쟁하거나 전 세계 Summit Circuit 경주 행사에서 직접 경쟁 할 수 있습니다.
Standings: Check out the live leaderboard and latest race results
일본의 스즈카 트랙에서 영감을 얻은 Kumo Torakku 서킷 레이싱에서 우승해 re:Invent로 가는 경비 지원을 받으세요. 후지산을 보며 도쿄의 거리를 달려 승리하세요. 그리고 포인트와 상금을 획득하세요. 그리고 AWS re:Invent 2019에서 열리는 AWS DeepRacer Championship Cup 출전권도 받으세요.
AWS Free Tier를 사용하면 최대 10 시간동안 훈련을 무료로 진행할 수 있습니다. 그러니까 AWS DeepRacer League에 아무런 비용을 들이지 않고 참여해 보세요.
Take a step inside the AWS DeepRacer League. Episode 1 follows the competition to Amsterdam, featuring developers of all skill levels hoping to qualify for a chance to win the Championship Cup at AWS re:Invent 2019. Tune in now to learn more about their strategy and how you can build and tune a model for a chance to win!
AWS DeepRacer League를 향해 한걸음 내디뎌 보세요. 에피소드 1은 암스테르담에서의 경기가 나옵니다. AWS re:Invent 2019 에서 열리는 Championship Cup에서 우승하는 것을 목표로 하는 다양한 수준의 개발자들이 나옵니다. 그들의 전략에 대해 배워 보세요 그리고 여러분은 우승을 위해 어떻게 모델을 만들어 갈 것인지 생각해 보세요.
Pick a race
개발자는 매월 공개 될 유명한 raceways에서 영감을 얻은 가상 트랙을 통해 Virtual Circuit에서 경쟁함으로써 테스트에 자신의 기술을 적용 할 수 있습니다. re:Invent 2019에서 열리는 AWS DeepRacer Championship Cup에 대한 경비와 경품을 얻기 위해 경쟁 할 수 있습니다.
20 개의 AWS Summit 중 아무 Summit이나 선택해서 직접 참가하세요. (참여하고 싶으면 여러 Summit에 참여하실 수 있습니다.) 워크샵을 통해 re:MARS 및 re:Invent 2019에서 모델을 어떻게 만들고 트레이닝 시킬 지에 대해 도움을 드릴 겁니다. 여러분이 트레이닝 시킨 모델을 집으로 가져 가실 수도 있습니다. 그리고 그 모델을 테스트하고 엑스포에서 그것을 가지고 경쟁하실 수 있습니다.
기계 학습을 처음 사용 하던지 기존 스킬을 기반으로 준비하든, 우리는 당신이 경주 준비를하는 것을 도울 수 있습니다. e- 러닝 수업 인 AWS DeepRacer : 강화 학습 학습에 의해 시작할 수 있습니다. 약 90 분 후에 보강 학습 (자율 차량 교육에 이상적인 기계 학습의 한 분야)의 기본 사항과 AWS DeepRacer - 트랙을 누빌 준비가되었습니다!
Summit Circuit 레이스 또는 Virtual Circuit 레이스에서 경쟁하십시오. 경쟁하는 레이스 수에는 제한이 없습니다. Virtual Circuit 또는 Summit Circuit 레이스 중 하나만이라도 1등을 하면 라스베가스에서 열리는 AWS re:Invent에서 열리는 AWS DeepRacer final round에 참가하는 경비를 지원받으실 수 있습니다. 각 레이스의 10등까지는 AWS DeepRacer 자동차를 받으실 수 있습니다.
각 레이스마다 참여하시면 포인트를 받게 됩니다. 2019년 말에 이 포인트들의 합계를 가지고 등수를 가립니다. 각 Circuit의 우승자에게는 re:Invent 2019의 AWS DeepRacer knockout(final) 라운드에 참가하는 경비를 지원 받으실 수 있습니다. 자세한 내용은 AWS DeepRacer 공식 규칙을 참조하십시오.
레이스에 참가하면 2019 AWS DeepRacer League Champion이 될 수 있습니다!
# Calculate 3 markers that are at varying distances away from the center line marker_1 = 0.1 * track_width marker_2 = 0.25 * track_width marker_3 = 0.5 * track_width
# Give higher reward if the car is closer to center line and vice versa if distance_from_center <= marker_1: reward = 1.0 elif distance_from_center <= marker_2: reward = 0.5 elif distance_from_center <= marker_3: reward = 0.1 else: reward = 1e-3 # likely crashed/ close to off track
return float(reward)
소스코드를 보니 트랙 중앙선 가까이 가면 점수(reward)를 더 많이 주는 간단한 로직이다.
30분 트레이닝 시키고 곧바로 Evaluate 결과는 조금 있다가…
이 첫번째 모델을 clone 해서 두번째 모델을 만들었다. 다 똑같고 reward function 함수만 내가 원하는 대로 조금 바꾸었다.
def reward_function(params): ''' Example of rewarding the agent to follow center line ''' reward=1e-3
if distance_from_center >=0.0 and distance_from_center <= 0.03: reward = 1.0
if not all_wheels_on_track: reward = -1 else: reward = params['progress']
# Steering penality threshold, change the number based on your action space setting ABS_STEERING_THRESHOLD = 15
# Penalize reward if the car is steering too much if steering > ABS_STEERING_THRESHOLD: reward *= 0.8
# add speed penalty if speed < 2.5: reward *=0.80
return float(reward)
이전에 썼던 디폴트 함수와는 다르게 몇가지 조건을 추가 했다. 일단 중앙선을 유지하면 좀 더 점수를 많이 주는 것은 좀 더 간단하게 만들었다. 이 부분은 이전에 훈련을 했으니까 이 정도로 해 주면 되지 않을까? 그리고 아무 바퀴라도 트랙 밖으로 나가면 -1을 하고 모두 트랙 안에 있으면 progress 만큼 reward를 주었다. Progress는 percentage of track completed 이다. 직진해서 결승선에 더 가까이 갈 수록 점수를 더 많이 따도록 했다. 이건 차가 빠꾸하지 않고 곧장 결승점으로 직진 하도록 만들기 위해 넣었다. 그리고 갑자기 핸들을 과하게 돌리면 차가 구르거나 트랙에서 이탈할 확률이 높으니 핸들을 너무 과하게 돌리면 점수가 깎이도록 했다. (15도 이상 핸들을 꺾으면 점수가 깎인다.) 그리고 속도도 너무 천천히 가면 점수를 깎는다.
속도 세팅이 최대 5로 만들어서 그 절반인 2.5 이하고 속도를 줄이면 점수가 깎인다.
이렇게 조건들을 추가하고 Training 시작. 이건 좀 복잡하니 트레이닝 시간을 1시간 주었다.
이 두개의 모델에 대한 결과는…
딱 보니 첫번째 디폴트 함수를 사용했을 때는 시간이 갈수록 결과가 좋게 나왔다. 그런데 두번째는 시간이 갈수록 실력이 높아지지는 않는 것 같다.
너무 조건이 여러개 들어가서 그런가?
생각해 보니 조건을 많이 넣는다고 좋은 것은 아닌것 같다. 일반적으로 코딩을 하다 보면 예외 상황을 만들지 않게 하기 위해 조건들을 아주 많이 주는 경향이 있는데 이 인공지능 쪽은 꼭 조건을 많이 줄 필요는 없을 것 같다.
앞으로 인공지능 쪽을 하다보면 일반 코딩에서의 버릇 중에 고칠 것들이 많을 것 같다. Evaluation 결과를 보면 두개의 차이가 별로 없다. 두 모델 모두 3번중 2번 완주 했고 완주시간도 비슷한 것 같다. 조건을 쪼금 더 준 Model 2 가 좀 더 낫긴 하네. (0.2 ~0.3 초 더 빠르다.) 다음은 곡선이 있는 다른 트랙으로 훈련을 시킬 계획이다.
그런데 곡선이 있는 트랙에서는 스피드가 무조건 빠르다고 좋은 건 아닌 것 같다. 내가 스피드를 5로 주었는데 Clone을 만들어서 할 때는 이 스피드를 조절하지 못하는 것 같다.
지난주 매일 1달러씩 부과 되던 NAT Gateways 서비스를 언제 세팅 했는지 몰랐었다.
여하튼 아마존 Support 서비스의 도움을 받아서 해당 서비스를 Delete 한 이후 요금은 더이상 부과 되지는 않았는데...
문제는 그걸 지우고 난 이후 Deepracer model을 생성하면 에러가 발생해서 일을 더이상 진행 할 수가 없었다.
도저히 안되서 새로운 account를 생성하고 Deepracer Model을 하나 생성했다.
그랬더니 모델과 더불어 NAT Gateways가 생성되더라.
이제 알았다. 처음 DeepRacer를 시작하기 위해 account resources를 생성할 때 이 NAT Gateway가 생성된다는 것을....
이 단계를 완료하면 NAT Gateways가 생성돼 매일 1달러씩 청구된다.
참고로 아래는 DeepRacer 모델을 하나 생성하면 청구되는 금액들이다.
* Data Transfer에서 $0.010 per GB - regional data transfer - in/out/between EC2 AZs or using elastic IPs or ELB 가 5기가가넘어서 5센트가 청구됐다.
* 그 다음 EC2에서 NAT Gateway가 생성과 연계해서 2.10 달러가 청구 됐다.
* 여기서 NAT Gateway는 매일 1달러씩 청구되게 돼 있다.
* 그 다음은 시뮬레이터 서비스인 RoboMaker 서비스에 대해 1.52달러가 청구됐고 SageMaker는 30센트가 청구됐다.
다른 서비스들은 사용하면 사용한 시간만큼만 내는데 NAT Gateway는 24시간 계속 돌아가기 때문에 매일 1불씩 내야 한다.
사용하지 않을 때 Stop 할 수도 없는 것 같다.
하루에 1달러이면 아무것도 아니라고 생각할 수 있지만... 아무것도 하지 않는데도 청구가 되니 속이 쓰렸다.
그래서 AWS Support Case를 하나 더 Raise 했다.
오늘 아침에 했는데 퇴근 할 때 쯤 답변이 오더라구.
다음은 답변 내용...
Hey there! This is Merlyn, your AWS Billing & Accounts specialist. I hope you are having a wonderful day! I understand you have some concerns about the DeepRacer pricing and you bill. Worry not as I'm here to help.
DeepRacer의 가격 책정과 청구서와 관련 우려가 있다는 것에대해 이해합니다. 걱정마세요. 제가 도와드릴테니까요.
With AWS DeepRacer console services, there are no charges for driving the AWS DeepRacer car or upfront charges. You will only pay for the AWS services you use. You will be billed separately by each of the AWS services used to provide the AWS DeepRacer console services such as creating and training your models, storing your models and logs, and evaluating them, including racing them in the virtual simulator or deploying them to their AWS DeepRacer car. You will see the bill for each service on your monthly billing statement.
AWS DeepRacer console services는 AWS DeepRacer 차량을 driving 하는데에 대한 가격 책정이나 선불로 어떠한 돈을 낼 필요가 없습니다. 당신이 AWS services를 사용한 것에 대한 요금만 지불 하시면 됩니다. AWS DeepRacer console services를 통해 사용하시는 AWS services들에 대해 각각의 사용료가 따로 청구 될 것입니다. 예를 들어 당신의 모델들을 생성하고 훈련하는 것 그리고 그 모델들이나 로그들을 저장하는 것, 그 모델들을 평가하는 작업, 가상 시뮬레이터에서 모델의 레이싱을 하거나 AWS DeepRacer car에 배치하는 일들에 대해 따로 사용료가 부과 됩니다.
Upon first use of AWS DeepRacer simulation in the AWS console, new customers will get 10 hours of Amazon SageMaker training, and 60 Simulation Unit (SU) hours for Amazon RoboMaker in the form of service credits; $6 for SageMaker, $24 for RoboMaker and $34 for NAT Gateway. The service credits are available for 30 days and applied at the end of the month. A typical AWS DeepRacer simulation uses up to six SUs per hour, thus you will be able to run a typical AWS DeepRacer simulation for up to 10 hours at no charge.
AWS console에서 AWS DeepRacer simulation을 처음 사용하는 새 고객에게는 10시간 동안 Amazon SageMaker training과 Amazon RoboMaker에 대해 60 Simulation Unit (SU) 시간에 대해 service credits이라는 형태로 무료 사용하실 수 있는 기회를 드립니다. SageMaker는 $6 , RoboMaker $24 그리고 NAT Gateway는 $34 입니다. 이 service credits은 30일간 사용 가능하고 그 달의 말에 적용됩니다. 일반적으로 AWS DeepRacer simulation 사용은 시간당 six SUs이므로 당신은 AWS DeepRacer simulation을 10시간 동안 무료로 사용하실 수 있습니다.
Please visit the link below for more pricing information:
This inquiry is a bit out of our scope. Here in the Billing and Accounts department we don’t handle technical questions. Please note that AWS unbundles technical support in order to lower the prices of the services themselves instead of billing it to all customers disregarding if they’d like to use it or not. However, if technical support is required, I would like to tell you we offer several Premium Support plans starting at just $29 a month, which enables you to speak to an engineer by email, chat, or phone depending on what support plan you choose.
당신의 요청 중 일부는 우리의 권한을 벗어나는 부분이 있습니다. 저희 Billing and Accounts department에서는 기술적인 질문은 다루지 않습니다. 저희는 모든 고객에게 비용 지불의 부담을 주지 않고 저렴한 가격을 유지하기 위해 기술 부문의 Support는 분리해서 다루고 있습니다. 만약 기술적인 support를 원하시면 월 $29 부터 시작되는 Premium Support plan 서비스들이 있습니다. 이 유료 기술 지원 서비스를 이용하시면 이메일, 채팅 그리고 전화등 여러분이 편한 방식으로 엔지니어들과 소통을 할 수 있습니다.
All Premium Support plans include direct access to Support Engineers, and these plans offer a tailored support experience that allows you to select the support level that best fits your needs. More information including pricing and how to sign up can be found here:
모든 Premium Support plan 들에는 Support Engineers과의 직접 access 권한이 포함되며 이러한 플랜들에서는 사용자의 요구 사항에 가장 적합한 서포트 레벨을 선택할 수 있는 맞춤 지원 서비스가 제공 됩니다. 가격 정책 및 가입 방법에 대한 자세한 정보는 여기를 참조하세요.
You can also search the Amazon Web Services Developer forums or post a new question. Our technical staff and the Amazon Web Services developer community participate in the forums regularly and might be able to help you. Generally, questions are responded to in about a day, though there is not a guaranteed response time.
그 외에 Amazon Web Services Developer forums에서 궁금한 내용을 검색해서 찾는 방법도 있습니다. 또한 새로운 질문을 이곳에 올릴 수도 있구요. Our technical staff와 Amazon Web Services developer community에 참여하는 개발자들로 부터 지원을 받으실 수도 있을 겁니다. 일반적으로 하루 정도 기다리면 답변을 받습니다. (응답 시간에 대한 보장은 없지만요.)
I certainly want you to have the best experience while using our services, so please feel free to let me know if you have any other concern in the meantime, it will be an honor to keep on assisting you. On behalf of Amazon Web Services, I wish you Happy Cloud Computing!
Best regards,
Merlyn N.
Amazon Web Services
==================================================================== Learn to work with the AWS Cloud. Get started with free online videos and self-paced labs at http://aws.amazon.com/training/ ====================================================================
Hands-on Exercise 1: Model Training Using AWS DeepRacer Console
This is the first of four exercises that you will encounter in this course. This first exercise guides you through building, training, and evaluating your first RL model using the AWS DeepRacer console. To access the instructions for three of these exercises,download and unzip this course package. For this particular exercise, find and open the relevant PDF file and follow the steps within to complete the exercise.
*Note:This exercise is designed to be completed in your AWS account. AWS DeepRacer is part of AWS Free Tier, so you can get started at no cost. For the first month after sign-up, you are offered a monthly free tier of 10 hours of Amazon SageMaker training and 60 simulation units of Amazon RoboMaker (enough to cover 10 hours of training). If you go beyond those free tier limits, you will accrue additional costs. For more information, see the AWS DeepRacer Pricing page.
AWSTemplateFormatVersion: "2010-09-09" Description: 'AWS DeepRacer: Driven by Reinforcement Learning' Parameters: SagemakerInstanceType: Description: 'Machine Learning instance type that should be used for Sagemaker Notebook' Type: String AllowedValues: - ml.t2.medium - ml.t2.large - ml.t2.xlarge - ml.t3.medium - ml.t3.large - ml.t3.xlarge - ml.m5.xlarge Default: ml.t3.medium CreateS3Bucket: Description: Create and use a bucket created via this template for model storage Default: True Type: String AllowedValues: - True - False ConstraintDescription: Must be defined at True|False. VPCCIDR: Description: 'CIDR Block for VPC (Do Not Edit)' Type: String Default: 10.96.0.0/16 PUBSUBNETA: Description: 'Public Subnet A (Do Not Edit)' Type: String Default: 10.96.6.0/24 PUBSUBNETB: Description: 'Public Subnet B (Do Not Edit)' Type: String Default: 10.96.7.0/24 PUBSUBNETC: Description: 'Public Subnet C (Do Not Edit)' Type: String Default: 10.96.8.0/24 PUBSUBNETD: Description: 'Public Subnet D (Do Not Edit)' Type: String Default: 10.96.9.0/24 S3PathPrefix: Type: String Description: 'Bootstrap resources prefix' Default: 'awsu-spl-dev/spl-227' S3ResourceBucket: Type: String Description: 'Bootstrap S3 Bucket' Default: 'aws-training' Conditions: CreateS3Bucket: !Equals [ !Ref CreateS3Bucket, True ] # NoCreateS3Bucket: !Equals [ !Ref CreateS3Bucket, False ] Resources:
# Defining the VPC Used for the sanbox ENV, and notebook instance VPC: Type: 'AWS::EC2::VPC' Properties: CidrBlock: !Ref VPCCIDR EnableDnsSupport: 'true' EnableDnsHostnames: 'true' Tags: - Key: Name Value: 'DeepRacer Sandbox' # There is a few calls made to public to download supporting resources InternetGateway: Type: 'AWS::EC2::InternetGateway' DependsOn: VPC Properties: Tags: - Key: Name Value: 'DeepRacer Sandbox IGW' # Attached this IGW to the sanbox VPC AttachGateway: Type: 'AWS::EC2::VPCGatewayAttachment' DependsOn: - VPC - InternetGateway Properties: VpcId: !Ref VPC InternetGatewayId: !Ref InternetGateway # Default setting in the notebook is to use Public IP address to communicate # between instances running the simulation, and the instances collecting and # processing. A NatGW could have been used with added costs, but would allow for # use of private IP address.
# Found in testing that not all ML instance types may not be deployed or avaliable # in all AZ's within a given region. We are using the newest instance family of T3 PublicSubnetA: Type: 'AWS::EC2::Subnet' DependsOn: VPC Properties: VpcId: !Ref VPC CidrBlock: !Ref PUBSUBNETA AvailabilityZone: !Select - '0' - !GetAZs '' Tags: - Key: Name Value: 'Deepracer Sandbox - Public Subnet - A' PublicSubnetB: Type: 'AWS::EC2::Subnet' DependsOn: VPC Properties: VpcId: !Ref VPC CidrBlock: !Ref PUBSUBNETB AvailabilityZone: !Select - '1' - !GetAZs '' Tags: - Key: Name Value: 'Deepracer Sandbox Public Subnet - B' PublicSubnetC: Type: 'AWS::EC2::Subnet' DependsOn: VPC Properties: VpcId: !Ref VPC CidrBlock: !Ref PUBSUBNETC AvailabilityZone: !Select - '2' - !GetAZs '' Tags: - Key: Name Value: 'Deepracer Sandbox Public Subnet - C' PublicSubnetD: Type: 'AWS::EC2::Subnet' DependsOn: VPC Properties: VpcId: !Ref VPC CidrBlock: !Ref PUBSUBNETD AvailabilityZone: !Select - '3' - !GetAZs '' Tags: - Key: Name Value: 'Deepracer Sandbox Public Subnet - D' # Define the Public Routing Table PublicRouteTable: Type: 'AWS::EC2::RouteTable' DependsOn: - VPC - AttachGateway Properties: VpcId: !Ref VPC Tags: - Key: Name Value: 'Deepracer Sandbox Public Routing Table' # And add in the default route to 0.0.0.0/0 PublicRouteIGW: Type: 'AWS::EC2::Route' DependsOn: - PublicRouteTable - InternetGateway Properties: RouteTableId: !Ref PublicRouteTable DestinationCidrBlock: 0.0.0.0/0 GatewayId: !Ref InternetGateway # Attach the routing table to each of the subnets PublicRouteTableAssociationA: Type: 'AWS::EC2::SubnetRouteTableAssociation' Properties: SubnetId: !Ref PublicSubnetA RouteTableId: !Ref PublicRouteTable PublicRouteTableAssociationB: Type: 'AWS::EC2::SubnetRouteTableAssociation' Properties: SubnetId: !Ref PublicSubnetB RouteTableId: !Ref PublicRouteTable PublicRouteTableAssociationC: Type: 'AWS::EC2::SubnetRouteTableAssociation' Properties: SubnetId: !Ref PublicSubnetC RouteTableId: !Ref PublicRouteTable PublicRouteTableAssociationD: Type: 'AWS::EC2::SubnetRouteTableAssociation' Properties: SubnetId: !Ref PublicSubnetD RouteTableId: !Ref PublicRouteTable # Define a S3 endpoint for all the S3 traffic during training S3Endpoint: Type: AWS::EC2::VPCEndpoint Properties: VpcId: !Ref VPC RouteTableIds: - !Ref PublicRouteTable ServiceName: !Join - '' - - com.amazonaws. - !Ref 'AWS::Region' - .s3 PolicyDocument: Version: 2012-10-17 Statement: - Effect: Allow Principal: '*' Action: - 's3:*' Resource: - '*' # This exercise is going to need a bucket to store any file generated from training # There is a conditions to evaluate if the PRAM is true, else this resource would # not be created. SandboxBucket: Type: 'AWS::S3::Bucket' DeletionPolicy: Retain Condition: CreateS3Bucket Properties: BucketName: Fn::Join: - "-" - - deepracer-trainingexercise - Ref: AWS::Region - Ref: AWS::AccountId # Sagemaker is going to be making calls to Robomaker to launch the sim, and # Sagemaker to launch the training insance. This requries AWS credentals. A # Principal of sagemaker and robomaker needs to be assiged as both service will # assuming this role. Default Sagemaker full access and s3 access is needed. SageMakerNotebookInstanceRole: Type: 'AWS::IAM::Role' Properties: AssumeRolePolicyDocument: Version: 2012-10-17 Statement: - Effect: Allow Principal: Service: - sagemaker.amazonaws.com - robomaker.amazonaws.com Action: - 'sts:AssumeRole' ManagedPolicyArns: - 'arn:aws:iam::aws:policy/AmazonSageMakerFullAccess' Path: / Policies: - PolicyName: DeepRacerPolicy PolicyDocument: Version: 2012-10-17 Statement: - Effect: Allow Action: [ 's3:*', 'iam:GetRole' ] Resource: '*' # This is how the notebook gets loaded on to sagemaker. There is a zip file with # with the needed files, and a second http call to pull down the notebook. # This is only done "OnCreate" - when the sagemaker instance is first deployed # You can can the script get run "OnStart" (when a sagemaker instance changes # from a stopped state to a running state). This would automaticlly update file # to be the latest form source, but could over write changes applied during # your testing SageMakerNotebookInstanceLifecycleConfig: Type: 'AWS::SageMaker::NotebookInstanceLifecycleConfig' Properties: # OnStart: # - Content: # Fn::Base64: # #!/bin/bash # !Sub | # cd SageMaker # chown ec2-user:ec2-user -R /home/ec2-user/SageMaker
OnCreate: - Content: Fn::Base64: !Sub | cd SageMaker curl -O https://us-west-2-${S3ResourceBucket}.s3.amazonaws.com/${S3PathPrefix}/scripts/rl_deepracer_robomaker_coach.ipynb curl -O https://us-west-2-${S3ResourceBucket}.s3.amazonaws.com/${S3PathPrefix}/scripts/rl_deepracer_robomaker_coach.zip unzip rl_deepracer_robomaker_coach.zip chown ec2-user:ec2-user -R /home/ec2-user/SageMaker # Security Group for sagemaker instance running in this VPC SagemakerInstanceSecurityGroup: Type: AWS::EC2::SecurityGroup Properties: GroupDescription: Sagemaker Security Group VpcId: !Ref VPC SecurityGroupIngress: - IpProtocol: tcp FromPort: 1 ToPort: 65535 CidrIp: !Ref VPCCIDR - IpProtocol: udp FromPort: 1 ToPort: 65535 CidrIp: !Ref VPCCIDR SecurityGroupEgress: - IpProtocol: tcp FromPort: 1 ToPort: 65535 CidrIp: !Ref VPCCIDR - IpProtocol: udp FromPort: 1 ToPort: 65535 CidrIp: !Ref VPCCIDR # Creating the Sagemaker Notebook Instance SageMakerNotebookInstance: Type: 'AWS::SageMaker::NotebookInstance' Properties: #NotebookInstanceName: 'DeepracerSagemakerSandbox' NotebookInstanceName: !Join ["-", ["DeepRacerSagemakerSandbox", !Ref "AWS::StackName"]] SecurityGroupIds: - !GetAtt - SagemakerInstanceSecurityGroup - GroupId InstanceType: !Ref SagemakerInstanceType SubnetId: !Ref PublicSubnetA Tags: - Key: Name Value: 'DeepRacer Sandbox' LifecycleConfigName: !GetAtt - SageMakerNotebookInstanceLifecycleConfig - NotebookInstanceLifecycleConfigName RoleArn: !GetAtt - SageMakerNotebookInstanceRole - Arn Outputs: # Display the name of the bucekt that was created from this CFN Stack ModelBucket: Condition: CreateS3Bucket Value: !Ref SandboxBucket # URL to get to the Sagemaker UI, and find the Jupyter button. SagemakerNotebook: Value: !Sub | https://console.aws.amazon.com/sagemaker/home?region=${AWS::Region}#/notebook-instances/${SageMakerNotebookInstance.NotebookInstanceName}
Elastic Compute Cloud 를 expand 해서 봤더니 NAT Gateway hour 부분 액수가 증가했다.
이제 어디서 부과 되는지 알아냈으니 그걸 de-activate 시키면 될 일.
EC2 - Elastic IPs로 갔더니 NAT Gateway 가 하나 running 중이었다.
내가 이걸 수요일에 만들었었나?
전혀 기억이 안나지만 어쨌든 내가 만들었으니 생겼겠지....
그래서 그걸 선택하고 Detach를 시도했다.
근데 이런 에러메세지만 나오고 Detach는 거절 당했다.
"Error detaching network interface eni--------------: You are not allowed to manage 'ela-attach' attachments."
암만 시도해도 안되길래 AWS Support center에 질문을 올렸다.
월요일 아침에 바로 답장이 왔다.
아마존은 배달만 빠른줄 알았더니 문의에 대한 응답도 1일 배송이었다.
Hi there,
I hope you are having a great day.
I understand that you have some charges for NAT Gateway on your AWS account, I'm happy to assist.
1. How can I stop the service?
You can delete a NAT gateway using the Amazon VPC console by following the steps below.
1. Log into your Amazon VPC console: https://console.aws.amazon.com/vpc/ 2. In the navigation pane, choose NAT Gateways. 3. Select the NAT gateway, and then choose Delete NAT Gateway. 4. In the confirmation dialog box, choose Delete NAT Gateway.
After you've deleted a NAT gateway, its entry remains visible in the Amazon VPC console for a short while (usually an hour), after which it's automatically removed. You cannot remove this entry yourself.
More information on NAT Gateway can be found here:
It's important to note that deleting a NAT gateway disassociates its Elastic IP address, but does not release that Elastic IP address from your AWS account. You will incur charges for any Elastic IP address that remains allocated to your account, but is not associated with an instance.
We recommend you release a disassociated Elastic IP address from your account if you no longer need it. To release an Elastic IP address:
1. Open the Amazon VPC console at: https://console.aws.amazon.com/vpc/ . 2. In the navigation panel, choose Elastic IPs. 3. Select the Elastic IP address you wish to release, choose Actions, and then choose Release addresses. 4. When prompted, choose Release.
To learn more about Elastic IP addresses, please follow the link below:
2. When (and How) did I created the service? (I don't remember when and how I created it.
After reviewing your AWS account, I could see that this Nat Gateway was created on May 15, 2019. I've added a report so you can see that the charges started on May 15th.
We do not have the information on how it was created but this documentation can be helpful to you.
(If you are connecting by federation, log in before following the link.)
*Please note: this e-mail was sent from an address that cannot accept incoming e-mail. Please use the link above if you need to contact us again about this same issue.
==================================================================== Learn to work with the AWS Cloud. Get started with free online videos and self-paced labs at http://aws.amazon.com/training/ ====================================================================
Amazon Web Services, Inc. and affiliates
봤더니 NAT Gateway는 EC2가 아니라 VPC로 가서 Delete를 해 줘야 하는 거였다.
VPC - NAT Gateways - 선택 후 Delete
간단했다.
메일에서 추천한대로 Elastics IPs 로 가서 해당 Elastic IP address를 release 했다.
이걸로 사용도 하지 않는 Paid AWS Service 때문에 요금이 부과 되는 상황은 중지 시킨거 같다.
(RL)Reinforcement learning (RL)은 기계 학습 machine learning의 한 유형으로, agent가 원하는 임무를 어떻게 수행할 것인가를 배우기 위해 주어진 환경environment 을 탐구하는 것입니다. 좋은 결과를 위한 액션을 취하고 나쁜 결과를 초래하는 액션은 회피함으로서 원하는 임무에 최적화 된 행동을 학습하게 됩니다.
reinforcement learning 모델은 경험을 통해 배우며 시간이 지남에 따라 어떤 행동이 최상의 보상reward으로 이어질지 식별 할 수 있게 됩니다.
다른 타입의 machine learning
Supervised learning
예제 중심 교육 - 주어진 입력들에 대해 알려진 출력들에 대한 레이블된 데이터를 가지고 이 모델은 새로운 입력에 대한 출력을 예측하도록 훈련됩니다.
Unsupervised learning
추론 기반 교육 - 알려진 출력이 없는 레이블된 데이터를 가지고 이 모델은 입력 데이터 내의 관련 구조 또는 유사한 패턴을 식별하도록 훈련됩니다.
How does AWS DeepRacer learn to drive by itself?
In reinforcement learning, anagentinteracts with anenvironmentwith an objective to maximize its totalreward.
The agent takes anactionbased on the environmentstateand the environment returns the reward and the next state. The agent learns from trial and error, initially taking random actions and over time identifying the actions that lead to long-term rewards.
Let's explore these ideas and how they relate to AWS DeepRacer.
reinforcement learning에서 agent는 환경environment과 상호 작용하여 총 보상reward을 최대화합니다.
agent는 환경environment 상태state에 따라 조치를 취하고 환경environment은 보상reward과 다음 상태state를 반환합니다. 에이전트agent는 초기에 무작위로 행동을 취하고 시간이 지남에 따라 장기 보상long-term rewards으로 이어지는 행동을 식별함으로써 이러한 시행 착오를 통해 학습합니다.
이런 각각의 개념들을 살펴보고 AWS DeepRacer와의 관계에 대해서도 알아보겠습니다.
Agent
Theagentsimulates the AWS DeepRacer vehicle in the simulation for training. More specifically, it embodies the neural network that controls the vehicle, taking inputs and deciding actions.
에이전트agent는 훈련을 하기 위해 시뮬레이션에서 AWS DeepRacer 차량을 시뮬레이트 합니다. 보다 구체적으로, 차량을 제어하고 입력을 취하고 행동을 결정하는 신경망을 구현합니다.
Environment
Theenvironmentcontains a track that defines where the agent can go and what state it can be in. The agent explores the envrionment to collect data to train the underlying neural network.
환경environment 은 에이전트agent 가 어디로 갈 수 있고 어떤 상태에 놓일 수 있는 지에 대해 정의하는 트랙으로 구성됩니다. 에이전트는 기본 신경 네트워크를 훈련하기 위해 주어진 환경(트랙)을 탐색하면서 데이터를 수집합니다.
State
Astaterepresents a snapshot of the environment the agent is in at a point in time.
For AWS DeepRacer, a state is an image captured by the front-facing camera on the vehicle.
상태state는 특정 시점에 에이전트가있는 환경의 스냅 샷을 가리킵니다. AWS DeepRacer의 경우 상태는 차량의 전면 카메라가 캡처 한 이미지입니다.
Action
Anactionis a move made by the agent in the current state. For AWS DeepRacer, an action corresponds to a move at a particular speed and steering angle.
동작action은 현재 상태에서 agent가 수행 한 동작입니다. AWS DeepRacer의 경우 동작은 특정 속도 및 방향(핸들) 각도 등의 움직임과 관계된 액션을 말합니다.
Reward
Therewardis the score given as feedback to the agent when it takes an action in a given state.
In training the AWS DeepRacer model, the reward is returned by areward function. In general, you define or supply a reward function to specify what is desirable or undesirable action for the agent to take in a given state.
보상reward은 어떤 주어진 상태에서 action를 취했을 때 agent에게 피드백으로 주어지는 점수입니다.
AWS DeepRacer 모델을 교육 할 때 보상reward은 reward function에 의해 반환됩니다. 일반적으로 agent 가 주어진 상태에서 취할 수있는 바람직하거나 바람직하지 않은 작업을 지정하기 위해 보상 기능을 정의하거나 제공합니다.
훈련은 반복적 인 과정입니다. 시뮬레이터에서 에이전트는 환경을 탐색하고 경험을 쌓습니다. 수집 된 경험은 신경망을 주기적으로 업데이트하는 데 사용되며 업데이트 된 모델은 더 많은 경험을 생성하는 데 사용됩니다.
AWS DeepRacer를 사용하여 자율 운전을 위한 차량을 교육합니다. 교육 과정을 시각화하는 것이 까다로울 수 있으므로 간단한 예를 살펴 보겠습니다.
이 예에서는 출발점에서 결승점까지 차량이 최단 경로로 갈 수 있도록 훈련시키려고 합니다.
우리는 환경environment 을 사각형 격자로 단순화했습니다. 각 사각형은 개별 상태를 나타내며, 목표 방향으로 차량을 위 또는 아래로 움직일 수 있습니다.
grid 내의 각 격자(사각형)마다 점수를 할당 할 수 있습니다. 그럼으로서 어떤 행동에 인센티브를 줄지 결정할 수 있게 됩니다.
여기서 우리는 트랙의 가장자리에있는 사각형들을 "정지 상태 stop states"로 지정하여 차량이 트랙에서 벗어 났음을 알립니다.
우리가 트랙의 중심을 주행하는 법을 배울 수 있게 하도록 차량에 인센티브를 부여하기를 원하기 때문에 센터 라인의 사각형에 높은 보상reward 을 설정하고 다른 곳에는 보다 낮은 보상reward을 설정합니다.
An episode
In reinforcement training, the vehicle will start by exploring the grid until it moves out of bounds or reaches the destination.
As it drives around, the vehicle accumulates rewards from the scores we defined. This process is called anepisode.
In this episode, the vehicle accumulates a total reward of2.2before reaching a stop state.
reinforcement training은 여기서 차량이 grid 를 탐색하기 시작하여 경계를 벗어나거나 목적지까지 도달하면 일단락 하게 됩니다.
차량이 움직이게 되면 우리가 지정한 점수에 따라 rewards 가 점점 쌓이게 됩니다. 이러한 과정을 에피소드라고 합니다. 위 에피소드에서 이 차량이 stop state에 도달하기 까지 총 2.2 점의 reward 를 쌓았습니다.
Iteration
Reinforcement learning algorithms are trained by repeated optimization of cumulative rewards.
Reinforcement learning algorithms은 누적된 rewards를 최적화 (최대화) 하기 위해 반복해서 훈련하게 됩니다.
The model will learn which action (and then subsequent actions) will result in the highest cumulative reward on the way to the goal.
이 모델은 우리가 목표로 하는 것을 달성하기 위한 최대화된 누적보상의 결과를 가져 올 수 있게 하는 것이 어떤 행동 (그리고 후속 행동)인가를 배우게 됩니다.
Learning doesn’t just happen on the first go; it takes some iteration. First, the agent needs to explore and see where it can get the highest rewards, before it can exploit that knowledge.
학습은 한번에 완성되지 않습니다. 반복이 필요합니다. 첫째, 에이전트는 knowledge를 취하기 전에 가장 높은 보상을 받을 수 있는 것이 어떤 것인지를 탐색해야 합니다.
Exploration
As the agent gains more and more experience, it learns to stay on the central squares to get higher rewards.
If we plot the total reward from each episode, we can see how the model performs and improves over time.
agent 는 점점 더 많은 경험을 쌓게 됩니다. 그러는 와중에 더 많은 점수를 얻으려면 중앙 격자를 계속 따라가야 한다는 것을 배우게 될 것입니다.
각 에피소드별 보상 점수가 어떤지를 보면 이 모델이 어떤 퍼포먼스를 보여주고 있고 얼마나 개선되고 있는지를 알 수 있습니다.
Exploitation and Convergence
With more experience, the agent gets better and eventually is able to reach the destination reliably.
경험이 많아질 수록 agent 는 점점 더 나아지고 결국에는 목표에 도달할 수 있게 됩니다.
Depending on the exploration-exploitation strategy, the vehicle may still have a small probability of taking random actions to explore the environment.
이 exploration-exploitation strategy(탐사-탐사 전략)에 근거해서 차량은 환경을 탐사하기 위한 랜덤한 액션 (무작위 행동)을 할 가능성이 점점 더 줄어들게 됩니다.
AWS DeepRacer에서 보상 기능 reward function은 현재 상태를 설명하고 숫자 보상 값을 반환하는 특정 매개 변수가있는 Python 함수입니다.
보상 기능으로 전달 된 매개 변수는 트랙의 위치와 방향, 관측 된 속도, 조향 각도 등과 같은 차량 상태의 다양한 측면을 나타냅니다.
우리는이 매개 변수 중 몇 가지를 탐색하고 트랙 주변을 따라 차량을 묘사하는 방법을 살펴 보겠습니다.
Position on track
Heading
Waypoints
Track width
Distance from center line
All wheels on track
Speed
Steering angle
1. Position on track
The parametersxandydescribe the position of the vehicle in meters, measured from the lower-left corner of the environment.
매개 변수 x 및 y는 환경의 왼쪽 하단 모서리에서 측정 한 차량의 위치를 미터 단위로 나타냅니다.
2. Heading
Theheadingparameter describes the orientation of the vehicle in degrees, measured counter-clockwise from the X-axis of the coordinate system.
heading 매개 변수는 좌표계의 X 축에서 시계 반대 방향으로 측정 한 차량의 방향을 도 단위로 나타냅니다.
3. Waypoints
Thewaypointsparameter is an ordered list of milestones placed along the track center.
Each waypoint inwaypointsis a pair[x, y]of coordinates in meters, measured in the same coordinate system as the car's position.
waypoints 매개 변수는 트랙 센터를 따라 배치 된 마일스톤의 정렬 된 목록입니다. waypoints 안에있는 각 waypoints 는 자동차의 위치에 대해 좌표계에서 측정 된 미터 단위의 좌표 [x, y] 쌍입니다.
4. Track width
Thetrack_widthparameter is the width of the track in meters.
track_widthparameter는 미터로 된 트랙의 너비 입니다.
5. Distance from center line
Thedistance_from_centerparameter measures the displacement of the vehicle from the center of the track.
Theis_left_of_centerparameter is a boolean describing whether the vehicle is to the left of the center line of the track.
distance_from_center 매개 변수는 트랙의 중심에서 차량의 변위를 측정합니다. is_left_of_center 매개 변수는 차량이 트랙의 중심선 왼쪽에 있는지 여부를 나타내는 boolean 값입니다.
6. All wheels on track
Theall_wheels_on_trackparameter is a boolean (true / false) which is true if all four wheels of the vehicle are inside the track borders, and false if any wheel is outside the track.
all_wheels_on_track 매개 변수는 boolean (true / false)이며 차량의 네 바퀴가 모두 트랙 테두리 안에 있으면 true이고, 하나라도 바퀴가 밖에 있으면 false가 됩니다.
7. Speed
Thespeedparameter measures the observed speed of the vehicle, measured in meters per second.
speedparameter는 초속 몇미터인가를 나타내는 차량의 관측된 속도입니다.
8. Steering angle
Thesteering_angleparameter measures the steering angle of the vehicle, measured in degrees.
This value is negative if the vehicle is steering right, and positive if the vehicle is steering left.
steering_angle 매개 변수는 차량의 조향 각도를 도 단위로 측정합니다. 이 값은 차량이 오른쪽으로 조향하는 경우 음수이고 차량이 좌회전하는 경우 양수입니다.
With all these parameters at your disposal, you can define a reward function to incentivize whatever driving behavior you like.
Let's see a few examples of reward functions and how they use the parameters to determine a reward. The following three reward functions are available as examples in the AWS DeepRacer console so you can try them out and see how they behave, or submit them to the AWS DeepRacer League.
이러한 모든 매개 변수를 마음대로 활용하면 원하는 운전 행동에 대한 인센티브를 주는 보상 기능을 정의 할 수 있습니다.
보상 기능의 몇 가지 예와 보상을 결정하기 위해 매개 변수를 사용하는 방법을 살펴 보겠습니다. AWS DeepRacer 콘솔에서 다음과 같은 세 가지 보상 기능을 사용할 수 있으므로 이를 시험해보고 어떻게 동작하는지 보거나 AWS DeepRacer League에 제출할 수 있습니다.
이 예에서는 자동차가 궤도에 머무를 때 높은 보상을주고 차가 궤도 경계를 벗어나는 경우 페널티를줍니다.
이 예제에서는 all_wheels_on_track, distance_from_center 및 track_width 매개 변수를 사용하여 자동차가 트랙에 있는지 여부를 확인하고 높은 경우 보상을 제공합니다.
이 기능은 트랙에 머무르는 것 이외의 특정 종류의 행동에 대해 보상하지 않으므로이 기능으로 교육받은 agent 는 특정 행동으로 수렴하는 데 시간이 오래 걸릴 수 있습니다.
이 예제에서 우리는 트랙의 중심에서 차가 얼마나 떨어져 있는지 측정하고, 차가 중심선에 가까이 있으면 더 높은 보상을줍니다.
이 예제에서는 track_width 및 distance_from_center 매개 변수를 사용하고 트랙의 중심에서 차가 멀어 질수록 보상이 줄어 듭니다.
이 예는 보상의 대상이되는 운전 행동의 유형에 따라 다르므로이 기능을 사용하여 교육을받은 agent는 트랙을 잘 따라갈 수 있습니다. 그러나 모서리의 가속이나 제동과 같은 다른 행동을 배우기는 쉽지 않습니다.
대안의 전략 alternative strategy은 자동차가 운전하는 방식에 관계없이 각 단계에 대해 지속적인 보상을 제공하는 것입니다.
이 예제는 입력 매개 변수를 사용하지 않고 각 단계마다 상수 보상 1.0을 반환합니다.
에이전트의 유일한 인센티브는 트랙을 성공적으로 끝내는 것이고, 더 빠르게 운전하거나 특정 경로를 따라갈 인센티브가 없습니다. 그것은 불규칙하게 행동 할 수 있습니다.
그러나 보상 기능은 agent의 행동을 제한하지 않으므로 예상치 못한 전략과 행동을 탐색하여 오히려 실적이 좋을 수도 있습니다.
NumPy is the fundamental package for scientific computing with Python. It contains among other things:
a powerful N-dimensional array object
sophisticated (broadcasting) functions
tools for integrating C/C++ and Fortran code
useful linear algebra, Fourier transform, and random number capabilities
Besides its obvious scientific uses, NumPy can also be used as an efficient multi-dimensional container of generic data. Arbitrary data-types can be defined. This allows NumPy to seamlessly and speedily integrate with a wide variety of databases.
NumPy is licensed under theBSD license, enabling reuse with few restrictions.
Boto is the Amazon Web Services (AWS) SDK for Python. It enables Python developers to create, configure, and manage AWS services, such as EC2 and S3. Boto provides an easy to use, object-oriented API, as well as low-level access to AWS services.
Amazon ML supports three types of ML models: binary classification, multiclass classification, and regression. The type of model you should choose depends on the type of target that you want to predict.
Binary Classification Model
ML models for binary classification problems predict a binary outcome (one of two possible classes). To train binary classification models, Amazon ML uses the industry-standard learning algorithm known as logistic regression.
Examples of Binary Classification Problems
"Is this email spam or not spam?"
"Will the customer buy this product?"
"Is this product a book or a farm animal?"
"Is this review written by a customer or a robot?"
Multiclass Classification Model
ML models for multiclass classification problems allow you to generate predictions for multiple classes (predict one of more than two outcomes). For training multiclass models, Amazon ML uses the industry-standard learning algorithm known as multinomial logistic regression.
Examples of Multiclass Problems
"Is this product a book, movie, or clothing?"
"Is this movie a romantic comedy, documentary, or thriller?"
"Which category of products is most interesting to this customer?"
Regression Model
ML models for regression problems predict a numeric value. For training regression models, Amazon ML uses the industry-standard learning algorithm known as linear regression.
Examples of Regression Problems
"What will the temperature be in Seattle tomorrow?"
Amazon ML은 이진수 분류, 멀티클래스 분류 및 회귀라는 세 가지 유형의 ML 모델을 지원합니다. 선택해야 하는 모델 유형은 예측하려는 목표의 유형에 따라 따릅니다.
이진 분류 모델
이진 분류 문제에 대한 ML 모델은 이진 결과(가능성이 있는 두 가지 클래스 중 하나)를 예측합니다. 이진수 분류 모델을 교육하기 위해 은 '로지스틱 회귀'로 알려진 업계 표준 학습 알고리즘을 사용합니다.
이진 분류 문제의 예
"이 이메일은 스팸입니까? 스팸이 아닙니까?"
"고객이 이 제품을 구입할 것입니까?"
"이 제품은 책입니까? 아니면 가축입니까?"
"이 리뷰는 고객이 작성합니까? 로봇이 작성합니까?"
멀티클래스 분류 모델
멀티클래스 분류 문제에 대해 ML 모델을 사용하면 여러 클래스에 대한 예측을 생성할 수 있습니다(세 개 이상의 결과 중 하나를 예측). 멀티클래스 모델을 교육하기 위해 은 '다항 로지스틱 회귀'로 알려진 업계 표준 학습 알고리즘을 사용합니다.
멀티클래스 문제의 예
"이 제품은 책, 영화 또는 의류입니까?"
"이 영화는 로맨틱 코미디, 다큐멘터리 또는 스릴러입니까?"
"이 고객이 가장 관심을 갖는 제품 카테고리는 무엇입니까?"
회귀 모델
회귀 문제에 대해 ML 모델은 숫자 값을 예측합니다. 회귀 모델을 교육하기 위해 은 '선형 회귀'로 알려진 업계 표준 학습 알고리즘을 사용합니다.
회귀 문제의 예
"내일 시애틀의 기온은 어떨까요?"
"이 제품의 판매량이 얼마나 될까요?"
"이 집의 매매 가격이 얼마나 될까요?"
- Unsupervised Learning : Only data, Clustering Algorithm : Dimensionality Reduction Group words that are used in similar context or have similar meaning
- Reinforcement Learning Decision Making under uncertainty Autonomous Driving Games Reinforcement uses Reward Functions to reward correct decision and punish incorrect decision
Reinforcement learning (RL) is a machine learning technique that attempts to learn a strategy, called a policy, that optimizes an objective for an agent acting in an environment. For example, the agent might be a robot, the environment might be a maze, and the goal might be to successfully navigate the maze in the smallest amount of time. In RL, the agent takes an action, observes the state of the environment, and gets a reward based on the value of the current state of the environment. The goal is to maximize the long-term reward that the agent receives as a result of its actions. RL is well-suited for solving problems where an agent can make autonomous decisions.
RL is well-suited for solving large, complex problems. For example, supply chain management, HVAC systems, industrial robotics, game artificial intelligence, dialog systems, and autonomous vehicles. Because RL models learn by a continuous process of receiving rewards and punishments for every action taken by the agent, it is possible to train systems to make decisions under uncertainty and in dynamic environments.
Markov Decision Process (MDP)
RL is based on models called Markov Decision Processes (MDPs). An MDP consists of a series of time steps. Each time step consists of the following:
Environment
Defines the space in which the RL model operates. This can be either a real-world environment or a simulator. For example, if you train a physical autonomous vehicle on a physical road, that would be a real-world environment. If you train a computer program that models an autonomous vehicle driving on a road, that would be a simulator.
State
Specifies all information about the environment and past steps that is relevant to the future. For example, in an RL model in which a robot can move in any direction at any time step, then the position of the robot at the current time step is the state, because if we know where the robot is, it isn't necessary to know the steps it took to get there.
Action
What the agent does. For example, the robot takes a step forward.
Reward
A number that represents the value of the state that resulted from the last action that the agent took. For example, if the goal is for a robot to find treasure, the reward for finding treasure might be 5, and the reward for not finding treasure might be 0. The RL model attempts to find a strategy that optimizes the cumulative reward over the long term. This strategy is called apolicy.
Observation
Information about the state of the environment that is available to the agent at each step. This might be the entire state, or it might be just a part of the state. For example, the agent in a chess-playing model would be able to observe the entire state of the board at any step, but a robot in a maze might only be able to observe a small portion of the maze that it currently occupies.
Typically, training in RL consists of manyepisodes. An episode consists of all of the time steps in an MDP from the initial state until the environment reaches the terminal state.
Key Features of Amazon SageMaker RL
To train RL models in Amazon SageMaker RL, use the following components:
A deep learning (DL) framework. Currently, Amazon SageMaker supports RL in TensorFlow and Apache MXNet.
An RL toolkit. An RL toolkit manages the interaction between the agent and the environment, and provides a wide selection of state of the art RL algorithms. Amazon SageMaker supports the Intel Coach and Ray RLlib toolkits. For information about Intel Coach, seehttps://nervanasystems.github.io/coach/. For information about Ray RLlib, seehttps://ray.readthedocs.io/en/latest/rllib.html.
An RL environment. You can use custom environments, open-source environments, or commercial environments. For information, seeRL Environments in Amazon SageMaker.
The following diagram shows the RL components that are supported in Amazon SageMaker RL.
Amazon SageMaker RL을 사용한 강화 학습
강화 학습(RL)은 환경에서 작동하는 에이전트에 대한 목표를 최적화하는 전략(정책이라고 함)을 배우려고 시도하는 기계 학습 기법입니다. 예를 들어, 에이전트는 로봇, 환경은 미로, 목표는 최단시간 내에 미로를 성공적으로 탈출하는 것일 수 있습니다. RL에서 에이전트는 행동을 취하고, 환경의 상태를 관찰하고, 환경의 현재 상태 값에 따라 보상을 받습니다. 목표는 행동의 결과로 에이전트가 받는 장기 보상을 극대화하는 것입니다. RL은 에이전트가 자율 의사결정을 내릴 수 있는 문제를 해결하는 데 매우 적합합니다.
RL은 크고 복잡한 문제를 해결하는 데 매우 적합합니다. 예를 들어, 공급망 관리, HVAC 시스템, 산업용 로봇, 게임 인공 지능, 음성 대화 시스템 및 자율 주행 차량 등이 있습니다. RL 모델은 에이전트가 취하는 모든 행동에 대해 보상과 처벌을 받는 연속 프로세스를 통해 학습하기 때문에 동적인 환경에서 불확실성이 존재할 때 시스템이 의사를 결정하도록 훈련할 수 있습니다.
마코프 의사결정 과정(MDP)
RL은 마코프 의사결정 과정(MDP)라는 모델을 기반으로 합니다. MDP는 일련의 시간 단계로 구성됩니다. 각 시간 단계는 다음과 같은 요소로 구성됩니다.
환경
RL 모델이 작동하는 공간을 정의합니다. 이러한 공간은 실제 환경 또는 시뮬레이터일 수 있습니다. 예를 들어, 실제 도로에서 자율 주행 차량을 훈련하는 경우는 환경이 실제 환경입니다. 도로 위를 주행하는 자율 주행 차량을 모델링하는 컴퓨터 프로그램을 훈련하는 경우에는 환경이 시뮬레이터입니다.
상태
환경에 대한 모든 정보와 미래와 관련된 과거의 모든 단계를 지정합니다. 예를 들어, 로봇이 언제든지 어떤 방향으로든 이동할 수 있는 RL 모델에서는 현재 시간 단계에서 로봇의 위치가 상태입니다. 로봇 위치를 알면 해당 위치에 도착하기 위해 어떤 단계를 수행했는지 알 필요가 없기 때문입니다.
작업
작업은 에이전트가 수행합니다. 예를 들어 로봇이 앞으로 나아갑니다.
보상
에이전트가 수행한 마지막 작업의 상태 값을 나타내는 숫자입니다. 예를 들어, 목표가 로봇이 보물을 찾도록 하는 것이라면 보물을 찾은 경우 보상이 5이고, 보물을 찾지 못한 경우에는 보상이 0일 수 있습니다. RL 모델은 장기간 누적된 보상을 최적화하는 전략을 찾으려고 합니다. 이러한 전략을정책이라고 합니다.
관측치
각 단계마다 에이전트가 사용할 수 있는 환경 상태에 대한 정보입니다. 전체 상태이거나 상태의 일부분일 수 있습니다. 예를 들어, 체스 시합 모델의 에이전트는 모든 단계에서 체스판의 전체 상태를 관찰할 수 있지만 미로 속의 로봇은 현재 마주하고 있는 미로의 작은 부분 밖에 관찰할 수 없습니다.
일반적으로 RL의 훈련은 많은에피소드로 구성됩니다. 에피소드는 초기 상태에서 환경이 최종 상태에 도달할 때까지 MDP의 모든 시간 단계로 구성됩니다.
Amazon SageMaker RL의 주요 기능
Amazon SageMaker RL에서 RL 모델을 훈련하려면 다음 구성 요소를 사용합니다.
딥 러닝(DL) 프레임워크. 현재, Amazon SageMaker는 TensorFlow 및 Apache MXNet에서 RL을 지원합니다.
다음 다이어그램은 Amazon SageMaker RL에서 지원되는 RL 구성 요소를 보여 줍니다.
- refer to the picture above -
* Data types * Data in Real Life : Numeric, Text, Categorical values * Categorical : Cartesian Transformation - Combine categorical features to form new features * Text Type : NGRAM, OSB Transformation, Stemming, Lower Case * Numeric Data :
The n-gram transformation takes a text variable as input and produces strings corresponding to sliding a window of (user-configurable) n words, generating outputs in the process. For example, consider the text string "I really enjoyed reading this book".
Specifying the n-gram transformation with window size=1 simply gives you all the individual words in that string:
You can request n-grams with a size ranging from 2-10 words. N-grams with size 1 are generated implicitly for all inputs whose type is marked as text in the data schema, so you do not have to ask for them. Finally, keep in mind that n-grams are generated by breaking the input data on whitespace characters. That means that, for example, punctuation characters will be considered a part of the word tokens: generating n-grams with a window of 2 for string "red, green, blue" will yield {"red,", "green,", "blue,", "red, green", "green, blue"}. You can use the punctuation remover processor (described later in this document) to remove the punctuation symbols if this is not what you want.
To compute n-grams of window size 3 for variable var1:
"ngram(var1, 3)"
Orthogonal Sparse Bigram (OSB) Transformation
The OSB transformation is intended to aid in text string analysis and is an alternative to the bi-gram transformation (n-gram with window size 2). OSBs are generated by sliding the window of size n over the text, and outputting every pair of words that includes the first word in the window.
To build each OSB, its constituent words are joined by the "_" (underscore) character, and every skipped token is indicated by adding another underscore into the OSB. Thus, the OSB encodes not just the tokens seen within a window, but also an indication of number of tokens skipped within that same window.
To illustrate, consider the string "The quick brown fox jumps over the lazy dog", and OSBs of size 4. The six four-word windows, and the last two shorter windows from the end of the string are shown in the following example, as well OSBs generated from each:
Window, {OSBs generated}
"The quick brown fox", {The_quick, The__brown, The___fox} "quick brown fox jumps", {quick_brown, quick__fox, quick___jumps} "brown fox jumps over", {brown_fox, brown__jumps, brown___over} "fox jumps over the", {fox_jumps, fox__over, fox___the} "jumps over the lazy", {jumps_over, jumps__the, jumps___lazy} "over the lazy dog", {over_the, over__lazy, over___dog} "the lazy dog", {the_lazy, the__dog} "lazy dog", {lazy_dog}
Orthogonal sparse bigrams are an alternative for n-grams that might work better in some situations. If your data has large text fields (10 or more words), experiment to see which works better. Note that what constitutes a large text field may vary depending on the situation. However, with larger text fields, OSBs have been empirically shown to uniquely represent the text due to the specialskipsymbol (the underscore).
You can request a window size of 2 to 10 for OSB transformations on input text variables.
To compute OSBs with window size 5 for variable var1:
"osb(var1, 5)"
Lowercase Transformation
The lowercase transformation processor converts text inputs to lowercase. For example, given the input "The Quick Brown Fox Jumps Over the Lazy Dog", the processor will output "the quick brown fox jumps over the lazy dog".
To apply lowercase transformation to the variable var1:
"lowercase(var1)"
Remove Punctuation Transformation
Amazon ML implicitly splits inputs marked as text in the data schema on whitespace. Punctuation in the string ends up either adjoining word tokens, or as separate tokens entirely, depending on the whitespace surrounding it. If this is undesirable, the punctuation remover transformation may be used to remove punctuation symbols from generated features. For example, given the string "Welcome to AML - please fasten your seat-belts!", the following set of tokens is implicitly generated:
Note that only the prefix and suffix punctuation marks are removed. Punctuations that appear in the middle of a token, e.g. the hyphen in "seat-belts", are not removed.
To apply punctuation removal to the variable var1:
"no_punct(var1)"
Quantile Binning Transformation
The quantile binning processor takes two inputs, a numerical variable and a parameter calledbin number, and outputs a categorical variable. The purpose is to discover non-linearity in the variable's distribution by grouping observed values together.
In many cases, the relationship between a numeric variable and the target is not linear (the numeric variable value does not increase or decrease monotonically with the target). In such cases, it might be useful to bin the numeric feature into a categorical feature representing different ranges of the numeric feature. Each categorical feature value (bin) can then be modeled as having its own linear relationship with the target. For example, let's say you know that the continuous numeric featureaccount_ageis not linearly correlated with likelihood to purchase a book. You can bin age into categorical features that might be able to capture the relationship with the target more accurately.
The quantile binning processor can be used to instruct Amazon ML to establish n bins of equal size based on the distribution of all input values of the age variable, and then to substitute each number with a text token containing the bin. The optimum number of bins for a numeric variable is dependent on characteristics of the variable and its relationship to the target, and this is best determined through experimentation. Amazon ML suggests the optimal bin number for a numeric feature based on data statistics in theSuggested Recipe.
You can request between 5 and 1000 quantile bins to be computed for any numeric input variable.
To following example shows how to compute and use 50 bins in place of numeric variable var1:
"quantile_bin(var1, 50)"
Normalization Transformation
The normalization transformer normalizes numeric variables to have a mean of zero and variance of one. Normalization of numeric variables can help the learning process if there are very large range differences between numeric variables because variables with the highest magnitude could dominate the ML model, no matter if the feature is informative with respect to the target or not.
To apply this transformation to numeric variable var1, add this to the recipe:
normalize(var1)
This transformer can also take a user defined group of numeric variables or the pre-defined group for all numeric variables (ALL_NUMERIC) as input:
normalize(ALL_NUMERIC)
Note
It isnotmandatory to use the normalization processor for numeric variables.
Cartesian Product Transformation
The Cartesian transformation generates permutations of two or more text or categorical input variables. This transformation is used when an interaction between variables is suspected. For example, consider the bank marketing dataset that is used in Tutorial: Using Amazon ML to Predict Responses to a Marketing Offer. Using this dataset, we would like to predict whether a person would respond positively to a bank promotion, based on the economic and demographic information. We might suspect that the person's job type is somewhat important (perhaps there is a correlation between being employed in certain fields and having the money available), and the highest level of education attained is also important. We might also have a deeper intuition that there is a strong signal in the interaction of these two variables—for example, that the promotion is particularly well-suited to customers who are entrepreneurs who earned a university degree.
The Cartesian product transformation takes categorical variables or text as input, and produces new features that capture the interaction between these input variables. Specifically, for each training example, it will create a combination of features, and add them as a standalone feature. For example, let's say our simplified input rows look like this:
target, education, job
0, university.degree, technician
0, high.school, services
1, university.degree, admin
If we specify that the Cartesian transformation is to be applied to the categorical variables education and job fields, the resultant feature education_job_interaction will look like this:
target, education_job_interaction
0, university.degree_technician
0, high.school_services
1, university.degree_admin
The Cartesian transformation is even more powerful when it comes to working on sequences of tokens, as is the case when one of its arguments is a text variable that is implicitly or explicitly split into tokens. For example, consider the task of classifying a book as being a textbook or not. Intuitively, we might think that there is something about the book's title that can tell us it is a textbook (certain words might occur more frequently in textbooks' titles), and we might also think that there is something about the book's binding that is predictive (textbooks are more likely to be hardcover), but it's really the combination of some words in the title and binding that is most predictive. For a real-world example, the following table shows the results of applying the Cartesian processor to the input variables binding and title:
TextbookTitleBindingCartesian product of no_punct(Title) and Binding
n-gram 변환은 텍스트 변수를 입력으로 가져오고 (사용자가 구성 가능한) n 단어 창을 슬라이딩하는 문자열을 생성하여 해당 프로세스에서 출력을 생성합니다. 예를 들어 "I really enjoyed reading this book"이라는 텍스트 문자열을 생각해 보겠습니다.
창 크기 = 1인 n-gram 변환을 지정하면 다음과 같이 해당 문자열에 있는 모든 개별 단어가 제공됩니다.
n-gram을 2~10개 단어 범위의 크기로 요청할 수 있습니다. 크기가 1인 n-gram은 데이터 스키마에서 텍스트로 표시된 유형의 모든 입력에 대해 암시적으로 생성되므로, 사용자가 요청할 필요가 없습니다. 마지막으로, n-gram은 공백 문자에 대한 입력 데이터를 분해하여 생성된다는 것을 기억해야 합니다. 예를 들어 구두점 문자는 다음과 같이 단어 토큰의 일부로 간주됩니다. 문자열 "red, green, blue"에 대해 2개의 창으로 n-gram을 생성하면 {"red,", "green,", "blue,", "red, green", "green, blue"}로 산출됩니다. 구두점 제거 프로세서(이 문서의 뒷부분에서 설명)를 사용하여 원하지 않는 구두점을 제거할 수 있습니다.
변수 var1에 대해 창 크기가 3인 n-gram을 계산하려면:
"ngram(var1, 3)"
Orthogonal Sparse Bigram(OSB) 변환
OSB 변환은 텍스트 문자열 분석을 보조하는 목적이며 bi-gram 변환(창 크기가 2인 n-gram)을 대체하는 변환입니다. OSB는 n 단어 크기의 창을 텍스트 위로 슬라이딩하고 창의 첫 번째 단어가 포함된 모든 단어 페어를 출력하여 생성됩니다.
각 OSB를 구축하기 위해 구성 단어는 "_"(밑줄) 문자로 결합되며, 건너뛴 모든 토큰은 OSB에 또 다른 밑줄을 추가하여 표시됩니다. 따라서 OSB는 창 내에 확인되는 토큰뿐만 아니라 동일한 창 내에서 건너뛴 토큰의 수도 표시합니다.
예를 들어 "The quick brown fox jumps over the lazy dog"라는 문자열과 크기가 4인 OSB를 생각해 보겠습니다. 여섯 개의 4 단어 창 및 문자열 끝에 있는 더 짧은 길이의 마지막 창은 다음 예제에서 각 창에서 생성된 OSB와 함께 표시됩니다.
창, {생성된 OSB}
"The quick brown fox", {The_quick, The__brown, The___fox} "quick brown fox jumps", {quick_brown, quick__fox, quick___jumps} "brown fox jumps over", {brown_fox, brown__jumps, brown___over} "fox jumps over the", {fox_jumps, fox__over, fox___the} "jumps over the lazy", {jumps_over, jumps__the, jumps___lazy} "over the lazy dog", {over_the, over__lazy, over___dog} "the lazy dog", {the_lazy, the__dog} "lazy dog", {lazy_dog}
Orthogonal Sparse Bigram은 일부 상황에서 효과적으로 작용할 수 있는 n-gram의 대체 변환입니다. 데이터에 큰 텍스트 필드(10개 이상의 단어)가 있는 경우 실험을 통해 효과적인 방식을 살펴봅니다. 큰 텍스트 필드를 구성하는 내용은 상황에 따라 다를 수 있음을 참고하십시오. 그러나 큰 텍스트 필드를 사용하는 OSB는 특수건너뛰기기호(밑줄)로 인해 텍스트를 고유하게 나타내는 것으로 실증되었습니다.
입력 텍스트 변수의 OSB 변환에 대해 2~10의 창 크기를 요청할 수 있습니다.
변수 var1에 대해 창 크기가 5인 OSB를 계산하려면:
"osb(var1, 5)"
소문자 변환
소문자 변환 프로세서는 텍스트 입력을 소문자로 변환합니다. 예를 들어 프로세서는 "The Quick Brown Fox Jumps Over the Lazy Dog"라는 입력을 "the quick brown fox jumps over the lazy dog"라고 출력합니다.
변수 var1에 소문자 변환을 적용하려면:
"lowercase(var1)"
구두점 변환 제거
Amazon ML은 공백이 있는 데이터 스키마의 텍스트로 표시된 입력을 암시적으로 분리합니다. 문자열의 구두점은 주변의 공백에 따라 인접한 단어 토큰으로 끝나거나 완전히 별개의 토큰으로 끝납니다. 이것이 바람직하지 않은 경우, 생성된 특성의 구두점을 제거하는 데 구두점 제거 변환을 사용할 수 있습니다. 예를 들어 "Welcome to AML - please fasten your seat-belts!"라는 문자열이 있으면 다음과 같은 토큰 집합이 암시적으로 생성됩니다.
접두사 및 접미사 구두점만 제거됩니다. 토큰 중간에 나타나는 구두점("seat-belts"에 있는 하이픈 등)은 제거되지 않습니다.
변수 var1에 구두점 변환을 적용하려면:
"no_punct(var1)"
Quartile binning 변환
Quantile binning 프로세서는 숫자 변수 및bin number라는 파라미터의 두 가지 입력을 사용해서 카테고리 변수를 출력합니다. 그 목적은 관측된 값을 그룹화하여 변수의 배포에서 비선형성을 발견하는 것입니다.
많은 경우에 숫자 변수와 대상 간의 관계는 선형적이지 않습니다(숫자 변수 값은 대상에 따라 점차 증가하거나 감소하지 않음). 이 경우 다양한 범위의 숫자 특성을 나타내는 카테고리 특성으로 숫자 특성을 비닝하는 것이 유용할 수 있습니다. 각 카테고리 특성 값(빈)은 대상과의 선형 관계를 갖도록 모델링될 수 있습니다. 예를 들어 연속적인 숫자 특성account_age가 책을 구입할 가능성과 선형적으로 관련이 없다는 것을 알고 있다고 가정해 보겠습니다. 사용자는 대상과의 관계를 더욱 정확하게 캡처할 수 있는 카테고리 특성으로 age를 비닝할 수 있습니다.
quantile binning 프로세서는 Amazon ML이 age 변수의 모든 입력 값 배포를 기반으로 동일한 크기의 빈 n개를 설정한 다음 각 번호를 빈을 포함하는 텍스트 토큰으로 대체하도록 하는 데 사용될 수 있습니다. 숫자 변수에 대한 최적의 빈 수는 변수의 특징과 대상과의 관계에 따라 달라지며, 이는 실험을 통해 가장 효과적으로 결정됩니다. Amazon ML은제안된 레시피의 데이터 통계를 기반으로 숫자 특성에 대한 최적의 빈 수를 제시합니다.
숫자 입력 변수에 대해 5~1000개의 quantile 빈을 계산하도록 요청할 수 있습니다.
다음 예제에서는 숫자 변수 var1 대신 50개의 빈을 계산하여 사용하는 방법을 보여 줍니다.
"quantile_bin(var1, 50)"
정규화 변환
정규화 변환기는 평균 값이 0이고 분산 값이 1인 숫자 변수를 정규화합니다. 특성이 대상과 관련하여 많은 정보가 되는지 여부와 관계없이 가장 큰 크기의 변수가 ML 모델의 대부분을 차지할 수 있기 때문에, 숫자 변수의 정규화는 숫자 변수 사이에 범위 차이가 매우 큰 경우에 학습 프로세스에 도움이 될 수 있습니다.
숫자 변수 var1에 이 변환을 적용하려면 다음과 같이 레시피에 추가합니다.
normalize(var1)
이 변환기는 숫자 변수의 사용자 정의 그룹 또는 모든 숫자 변수(ALL_NUMERIC)의 사전 정의된 그룹을 입력으로 사용할 수 있습니다.
normalize(ALL_NUMERIC)
참고
숫자 변수에 정규화 프로세서를 사용하는 것은 필수가아닙니다.
데카르트 곱 변환
데카르트 변환은 두 개 이상의 텍스트 또는 카테고리 입력 변수의 순열을 생성합니다. 이 변환은 변수 간에 상호 작용이 있다고 생각될 때 사용합니다. 예를 들어 자습서: Amazon ML을 사용한 마케팅 반응 예측에서 사용되는 은행 마케팅 데이터 집합을 고려해 보겠습니다. 이 데이터 집합을 사용하여 경제 및 인구 통계 정보를 기반으로 은행 홍보에 긍정적인 반응을 보이는지 예측하려고 합니다. 해당하는 사람의 직업 유형이 상당히 중요하다고 생각할 수 있으며(특정 분야에서 근무하는 것과 사용 가능한 비용 간에 상관 관계가 있을 수 있음), 가장 높은 수준의 교육을 받은 사실도 중요하다고 간주할 수 있습니다. 또한 이 두 변수의 상호 작용에 강력한 신호가 있다는 것도 알 수 있습니다. 예를 들어 특히 학사 학위를 취득한 사업가 고객에게 홍보하는 것이 적합합니다.
데카르트 곱 변환은 카테고리 변수 또는 텍스트를 입력으로 사용하고, 이러한 입력 변수 간의 상호 작용을 캡처하는 새로운 특성을 생성합니다. 특히, 각 교육 예제에 대해 특성의 조합을 생성하고 이를 독립적인 특성으로 추가합니다. 예를 들어 다음과 같은 간소화된 입력 행이 있다고 가정하겠습니다.
target, education, job
0, university.degree, technician
0, high.school, services
1, university.degree, admin
데카르트 변환이 카테고리 변수 education 및 job 필드에 적용되도록 지정하면 결과적으로 다음과 같은 education_job_interaction 특성을 볼 수 있습니다.
target, education_job_interaction
0, university.degree_technician
0, high.school_services
1, university.degree_admin
데카르트 변환은 인수 중 하나가 암시적으로 또는 명시적으로 토큰으로 분리되는 텍스트 변수 같은 토큰 시퀀스 작업에서 더욱 강력하게 작용합니다. 예를 들어 책을 교과서로 분류할지 여부에 대한 작업을 생각해 보겠습니다. 직관적으로, 교과서라고 말할 수 있는 책의 제목에 대해 생각할 수 있으며(특정 단어는 교과서의 제목에서 더 자주 나타날 수 있음), 책의 표지에 대해 예측 가능한 점이 있다고 생각할 수도 있지만(교과서는 하드 커버가 될 가능성이 더 큼), 실제로 제목과 표지에 있는 단어의 조합이야말로 가장 예측 가능합니다. 실제 사례에서 다음 표는 데카르트 프로세서를 입력 변수 표지 및 제목에 적용한 결과를 보여 줍니다.
QUESTION: 1
A user is running a MySQL RDS instance. The user wil not use the DB for the
next 3 months.
How can the user save costs?
A. Pause the RDS activities from CLI until it is required in the future
B. Stop the RDS instance
C. Create a snapshot of RDS to launch in the future and terminate the instance
now
D. Change the instance size to micro
Answer(s): C Explanation:
The RDS instances unlike the AWS EBS backed instances cannot be stopped or
paused. The
user needs to take the final snapshot, terminate the instance and launch a new
instance in the
future from that snapshot. Reference:
http://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/Overview.BackingUpAndRestoring
AmazonR DSInstances.html
QUESTION: 2
In DynamoDB, if you create a table and request 10 units of write capacity and
200 units of read
capacity of provisioned throughput, how much would you be charged in US East
(Northern
Virginia) Region?
A. $0.05 per hour
B. $0.10 per hour
C. $0.03 per hour
D. $0.15 per hour
Answer(s): A Explanation:
To understand pricing in DynamoDB, consider the following example. If you
create a table and
request 10 units of write capacity and 200 units of read capacity of
provisioned throughput, you
would be charged:
$0.01 + (4 x $0.01) = $0.05 per hour Reference: http://aws.amazon.com/dynamodb/pricing/
QUESTION: 3
You have been doing a lot of testing of your VPC Network by deliberately
failing EC2 instances
to test whether instances are failing over properly. Your customer who wil be
paying the AWS
bil for all this asks you if he being charged for all these instances. You try
to explain to him how
the bil ing works on EC2 instances to the best of your knowledge. What would be
an appropriate
response to give to the customer in regards to this?
A. Bil ing commences when Amazon EC2 AMI instance is completely up and bil ing
ends as
soon as the instance starts to shutdown.
B. Bil ing commences when Amazon EC2 initiates the boot sequence of an AMI
instance and
bil ing ends when the instance shuts down.
C. Bil ing only commences only after 1 hour of uptime and bil ing ends when the
instance terminates.
D. Bil ing commences when Amazon EC2 initiates the boot sequence of an AMI
instance and
bil ing ends as soon as the instance starts to shutdown.
Answer(s): B Explanation:
Bil ing commences when Amazon EC2 initiates the boot sequence of an AMI
instance. Bil ing
ends when the instance shuts down, which could occur through a web services
command, by
running "shutdown -h", or through instance failure. Reference:
http://aws.amazon.com/ec2/faqs/#Bil ing
A. True. For specific availability zones only.
B. False
C. True. For specific regions only
D. True. For all regions
Answer(s): D Explanation:
You can configure your load balancer in ELB (Elastic Load Balancing) to use a
SSL certificate in
order to improve your system security.The load balancer uses the certificate to
terminate and
then decrypt requests before sending them to the back-end instances. Elastic
Load Balancing
uses AWS Identity and Access Management (IAM) to upload your certificate to your
load
balancer. Reference:
http://docs.aws.amazon.com/ElasticLoadBalancing/latest/DeveloperGuide/US_SettingUpLoadB
alancerH TTPS.html
QUESTION: 5
A user has launched five instances with ELB. How can the user add the sixth EC2
instance to
ELB?
A. The user can add the sixth instance on the fly.
B. The user must stop the ELB and add the sixth instance.
C. The user can add the instance and change the ELB config file.
D. The ELB can only have a maximum of five instances.
Answer(s): A Explanation:
Elastic Load Balancing automatically distributes incoming traffic across
multiple EC2 instances.
You create a load balancer and register instances with the load balancer in one
or more
Availability Zones. The load balancer serves as a single point of contact for
clients. This enables
you to increase the availability of your application. You can add and remove
EC2 instances from
your load balancer as your needs change, without disrupting the overall flow of
information. Reference:
http://docs.aws.amazon.com/ElasticLoadBalancing/latest/DeveloperGuide/SvcIntro.html
QUESTION: 6
An organization has 500 employees. The organization wants to set up AWS access
for each
department. Which of the below mentioned options is a possible solution?
A. Create IAM roles based on the permission and assign users to each role
B. Create IAM users and provide individual permission to each
C. Create IAM groups based on the permission and assign IAM users to the
groups
D. It is not possible to manage more than 100 IAM users with AWS
QUESTION: 7
How long can you keep your Amazon SQS messages in Amazon SQS queues?
A. From 120 secs up to 4 weeks
B. From 10 secs up to 7 days
C. From 60 secs up to 2 weeks
D. From 30 secs up to 1 week
Answer(s): C Explanation:
The SQS message retention period is configurable and can be set anywhere from 1
minute to 2
weeks. The default is 4 days and once the message retention limit is reached
your messages
wil be automatically deleted. The option for longer message retention provides
greater flexibility
to allow for longer intervals between message production and consumption. Reference: https://aws.amazon.com/sqs/faqs/
QUESTION: 8
In regard to DynamoDB, which of the following statements is correct?
A. An Item should have at least two value sets, a primary key and another
attribute.
B. An Item can have more than one attributes.
C. A primary key should be single-valued.
D. An attribute can have one or several other attributes.
Answer(s): B Explanation:
In Amazon DynamoDB, a database is a collection of tables. A table is a
collection of items and
each item is a collection of attributes. Reference:
http://docs.aws.amazon.com/amazondynamodb/latest/developerguide/DataModel.html
QUESTION: 9
Which one of the following statements is NOT an advantage of DyanamoDB being
built on Solid
State Drives:
A. serve high-scale request workloads
B. low request pricing
C. high I/O performance of WebApp on EC2 instance
D. low-latency response times
Answer(s): C Explanation:
In DynamoDB, SSDs help achieve design goals of predictable low-latency response
times for
storing and accessing data at any scale. The high I/O performance of SSDs also
enables to
serve high-scale request workloads cost efficiently, and to pass this
efficiency along in low
request pricing. Reference: http://aws.amazon.com/dynamodb/faqs/
QUESTION: 10
An organization has hosted an application on the EC2 instances. There will be
multiple users
connecting to the instance for setup and configuration of application. The
organization is
planning to implement certain security best practices. Which of the below
mentioned pointers
wil not help the organization achieve better security arrangement?
A. Apply the latest patch of OS and always keep it updated.
B. Al ow only IAM users to connect with the EC2 instances with their own secret
access key.
C. Disable the password based login for all the users. Al the users should use
their own keys to
connect with the instance securely.
D. Create a procedure to revoke the access rights of the individual user when
they are not
required to connect to EC2 instance anymore for the purpose of application
configuration.
Answer(s): B Explanation:
Since AWS is a public cloud any application hosted on EC2 is prone to hacker
attacks. It
becomes extremely important for a user to setup a proper security mechanism on
the EC2
instances. A few of the security measures are listed below:
Always keep the OS updated with the latest patch
Always create separate users with in OS if they need to connect with the EC2
instances, create
their keys and disable their password
Create a procedure using which the admin can revoke the access of the user when
the
business work on the EC2 instance is completed
Lock down unnecessary ports
Audit any proprietary applications that the user may be running on the EC2
instance Provide
temporary escalated privileges, such as sudo for users who need to perform
occasional
privileged tasks
The IAM is useful when users are required to work with AWS resources and
actions, such as
launching an instance. It is not useful to connect (RDP / SSH) with an
instance. Reference:http://aws.amazon.com/articles/1233/
QUESTION: 11
A user is planning to make a mobile game which can be played online or offline
and wil be
hosted on EC2. The user wants to ensure that if someone breaks the highest
score or they
achieve some milestone they can inform all their colleagues through email.
Which of the below
mentioned AWS services helps achieve this goal?
A. AWS Simple Workflow Service.
B. AWS Simple Queue Service.
C. Amazon Cognito
D. AWS Simple Email Service.
Answer(s): D Explanation:
Amazon Simple Email Service (Amazon SES) is a highly scalable and
cost-effective email-
sending service for businesses and developers. It integrates with other AWS
services, making it
easy to send emails from applications that are hosted on AWS. Reference: http://aws.amazon.com/ses/faqs/
QUESTION: 12
Which one of the following operations is NOT a DynamoDB operation?
A. BatchWriteItem
B. DescribeTable
C. BatchGetItem
D. BatchDeleteItem
QUESTION: 14
Regarding Amazon SNS, when you want to subscribe to a topic and receive
notifications to your
email, in the Protocol drop-down box, you should select _______.
A. Email
B. Message
C. SMTP
D. IMAP
Answer(s): A Explanation:
In Amazon SNS, when you want to subscribe to a topic and receive notifications
to your email,
select Email in the Protocol drop-down box. Enter an email address you can use
to receive the
notification in the Endpoint field. Reference: http://docs.aws.amazon.com/sns/latest/dg/SubscribeTopic.html
QUESTION: 15
In Amazon EC2, which of the following is the type of monitoring data for Amazon
EBS volumes
that is available automatically in 5-minute periods at no charge?
QUESTION: 16
In DynamoDB, to get a detailed listing of secondary indexes on a table, you can
use the _____
action.
A. DescribeTable
B. BatchGetItem
C. GetItem
D. TableName
Answer(s): A Explanation:
In DynamoDB, DescribeTable returns information about the table, including the
current status of
the table, when it was created, the primary key schema, and any indexes on the
table.
QUESTION: 17
A user has launched an EC2 instance. However, due to some reason the instance
was
terminated. If the user wants to find out the reason for termination, where can
he find the
details?
A. The user can get information from the AWS console, by checking the Instance
description
under the State transition reason label
B. The user can get information from the AWS console, by checking the Instance
description
under the Instance Termination reason label
C. The user can get information from the AWS console, by checking the Instance
description
under the Instance Status Change reason label
D. It is not possible to find the details after the instance is terminated
Answer(s): A Explanation:
An EC2 instance, once terminated, may be available in the AWS console for a
while after
termination. The user can find the details about the termination from the
description tab under
the label State transition reason. If the instance is stil running, there wil
be no reason listed. If
the user has explicitly stopped or terminated the instance, the reason wil be
"User initiated
shutdown". Reference:
http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/Using_InstanceStraightToTerminated.
html
QUESTION: 18
___________ is a task coordination and state management service for cloud
applications.
A. Amazon SES
B. Amazon SWF
C. Amazon FPS
D. Amazon SNS
Answer(s): B Explanation:
Amazon Simple Workflow (Amazon SWF) is a task coordination and state management
service
for cloud applications. With Amazon SWF, you can stop writing complex glue-code
and state
machinery and invest more in the business logic that makes your applications
unique. Reference: http://aws.amazon.com/swf/
QUESTION: 19
When you create a table with a hash-and-range key, you must define one or more
secondary
indexes on that table.
A. False, hash-range key is another name for secondary index
B. False, it is optional
C. True
D. False, when you have Hash-Range key you cannot define Secondary index
QUESTION: 20
A user is planning to create a structured database in the cloud. Which of the
below mentioned
AWS offerings help the user achieve the goal?
A. AWS DynamoDB
B. AWS RDS
C. AWS SimpleDB
D. AWS RSD
QUESTION: 21
A user has created a MySQL RDS instance with PIOPS. Which of the below
mentioned
statements wil help user understand the advantage of PIOPS?
A. The user can achieve additional dedicated capacity for the EBS I/O with an
enhanced RDS
option
B. It uses optimized EBS volumes and optimized configuration stacks
C. It provides a dedicated network bandwidth between EBS and RDS
D. It uses a standard EBS volume with optimized configuration the stacks
Answer(s): B Explanation:
RDS DB instance storage comes in two types: standard and provisioned IOPS.
Standard
storage is allocated on the Amazon EBS volumes and connected to the user's DB
instance.
Provisioned IOPS uses optimized EBS volumes and an optimized configuration
stack. It
provides additional, dedicated capacity for the EBS I/O. Reference: http://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/Welcome.html
QUESTION: 22
A user is accessing an EC2 instance on the SSH port for IP 10.20.30.40. Which
one is a secure
way to configure that the instance can be
accessed only from this IP?
A. In the security group, open port 22 for IP 10.20.30.40/0
B. In the security group, open port 22 for IP 10.20.30.40/32
C. In the security group, open port 22 for IP 10.20.30.40/24
D. In the security group, open port 22 for IP 10.20.30.40
Answer(s): B Explanation:
In AWS EC2, while configuring a security group, the user needs to specify the
IP address in
CIDR notation. The CIDR IP range 10.20.30.40/32 says it is for a single IP
10.20.30.40. If the
user specifies the IP as 10.20.30.40 only, the security group will not accept
and ask it in a CIRD
format. Reference: http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-network-
security.html
QUESTION: 23
When a user is detaching an EBS volume from a running instance and attaching it
to a new
instance, which of the below mentioned options should be followed to avoid file
system
damage?
A. Unmount the volume first
B. Stop all the I/O of the volume before processing
C. Take a snapshot of the volume before detaching
D. Force Detach the volume to ensure that all the data stays intact
Answer(s): A Explanation:
When a user is trying to detach an EBS volume, the user can either terminate
the instance or
explicitly remove the volume. It is a recommended practice to unmount the
volume first to avoid
any file system damage. Reference: http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ebs-detaching-
volume.html
QUESTION: 24
A user is planning to host a scalable dynamic web application on AWS. Which of
the services
may not be required by the user to achieve automated scalability?
A. CloudWatch
B. S3
C. AutoScaling
D. AWS EC2 instances
Answer(s): B Explanation:
The user can achieve automated scaling by launching different EC2 instances and
making them
a part of an ELB. Cloudwatch wil be used to monitor the resources and based on
the scaling
need it wil trigger policies. AutoScaling is then used to scale up or down the
instances. Reference:http://docs.aws.amazon.com/AutoScaling/latest/DeveloperGuide/WhatIsAutoScaling.html
QUESTION: 25
Which one of the following data types does Amazon DynamoDB not support?
A. Arrays
B. String
C. Binary
D. Number Set
Answer(s): A Explanation:
Amazon DynamoDB supports the following data types:
Scalar data types (like Number, String, and Binary)
Multi-valued types (like String Set, Number Set, and Binary Set). Reference:
http://docs.aws.amazon.com/amazondynamodb/latest/developerguide/DataModel.html#DataModel.Data
Types
QUESTION: 26
Regarding Amazon SNS, you can send notification messages to mobile devices
through any of
the following supported push notification services, EXCEPT:
A. Google Cloud Messaging for Android (GCM)
B. Apple Push Notification Service (APNS)
C. Amazon Device Messaging (ADM)
D. Microsoft Windows Mobile Messaging (MWMM)
Answer(s): D Explanation:
In Amazon SNS, you have the ability to send notification messages directly to
apps on mobile
devices. Notification messages sent to a mobile endpoint can appear in the
mobile app as
message alerts, badge updates, or even sound alerts. Microsoft Windows Mobile Messaging
(MWMM) doesn't exist and is not supported by Amazon SNS. Reference:
http://docs.aws.amazon.com/sns/latest/dg/SNSMobilePush.html
QUESTION: 27
A user plans to use RDS as a managed DB platform. Which of the below mentioned
features is
not supported by RDS?
A. Automated backup
B. Automated scaling to manage a higher load
C. Automated failure detection and recovery
D. Automated software patching
Answer(s): B Explanation: AWS RDS provides a managed DB platform, which offers
features, such as automated backup,
patch management, automated failure detection and recovery. The scaling is not
automated and
the user needs to plan it with a few clicks. Reference: http://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/Welcome.html
QUESTION: 28
A user has not enabled versioning on an S3 bucket. What wil be the version ID
of the object
inside that bucket?
A. 0
B. There wil be no version attached
C. Null
D. Blank
Answer(s): C Explanation:
S3 objects stored in the bucket before the user has set the versioning state
have a version ID of
nul . When the user enables versioning, the objects in the bucket do not change
and their ID
remains null. Reference:
http://docs.aws.amazon.com/AmazonS3/latest/dev/AddingObjectstoVersionSuspendedBuckets.
html
QUESTION: 29
A user has created a queue named "myqueue" with SQS. There are four
messages published
to queue which are not received by the consumer yet. If the user tries to
delete the queue, what
wil happen?
A. A user can never delete a queue manual y. AWS deletes it after 30 days of
inactivity on
queue
B. It will initiate the delete but wait for four days before deleting until all
messages are deleted
automatically.
C. It wil ask user to delete the messages first
D. It wil delete the queue
Answer(s): D Explanation:
SQS allows the user to move data between distributed components of applications
so they can
perform different tasks without losing messages or requiring each component to
be always
available. The user can delete a queue at any time, whether it is empty or not.
It is important to
note that queues retain messages for a set period of time. By default, a queue
retains
messages for four days. Reference:
http://docs.aws.amazon.com/AWSSimpleQueueService/latest/SQSDeveloperGuide/SQSConce
pts.html
QUESTION: 30
What happens if your application performs
more reads or writes than your provisioned capacity?
A. Nothing
B. requests above your provisioned capacity wil be performed but you wil
receive 400 error
codes.
C. requests above your provisioned capacity wil be performed but you will
receive 200 error
codes.
D. requests above your provisioned capacity wil be throttled and you wil
receive 400 error
codes.
Answer(s): D Explanation:
Speaking about DynamoDB, if your application performs more reads/second or
writes/second
than your table's provisioned throughput capacity allows, requests above your
provisioned
capacity wil be throttled and you wil receive 400 error codes. Reference:
http://docs.aws.amazon.com/amazondynamodb/latest/developerguide/ProvisionedThroughputIn
tro.html
QUESTION: 31
In relation to Amazon SQS, how can you ensure that messages are delivered in
order?
A. Increase the size of your queue
B. Send them with a timestamp
C. Give each message a unique id.
D. AWS cannot guarantee that you wil receive messages in the exact order you
sent them
Answer(s): D Explanation:
Amazon SQS makes a best effort to preserve order in messages, but due to the
distributed
nature of the queue, AWS cannot guarantee that you will receive messages in the
exact order
you sent them. You typically place sequencing information or timestamps in your
messages so
that you can reorder them upon receipt. Reference: https://aws.amazon.com/items/1343?externalID=1343
QUESTION: 32
An organization has launched two applications: one for blogging and one for ECM
on the same
AWS Linux EC2 instance running in the AWS VPC. The organization has attached
two private
IPs (primary and secondary) to the above mentioned instance. The organization
wants the
instance OS to recognize the secondary IP address. How can the organization
configure this?
A. Use the ec2-net-utility package which updates routing tables, uses DHCP to
refresh the
secondary IP and adds the network interface.
B. Use the ec2-net-utils package which wil configure an additional network
interface and update
the routing table
C. Use the ec2-ip-update package which can configure the network interface as
well as update
the secondary IP with DHCP.
D. Use the ec2-ip-utility package which can update the routing tables as well
as refresh the secondary IP using DHCP.
Answer(s): B Explanation:
A Virtual Private Cloud (VPC) is a virtual network dedicated to the user's AWS
account. It
enables the user to launch AWS resources into a virtual network that the user
has defined. With
VPC the user can specify multiple private IP addresses for his instances. The
number of
network interfaces and private IP addresses that a user can specify for an instance
depends on
the instance type. This scenario helps when the user wants to host multiple
websites on a single
EC2 instance. After the user has assigned a secondary private IP address to his
instance, he
needs to configure the operating system on that instance to recognize the
secondary private IP
address. For AWS Linux, the ec2-net-utils package can take care of this step.
It configures
additional network interfaces that the user can attach while the instance is
running, refreshes
secondary IP addresses during DHCP lease renewal, and updates the related
routing rules. Reference:
http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/MultipleIP.html
QUESTION: 33
What kind of service is provided by AWS DynamoDB?
A. Relational Database
B. NoSQL Database
C. Dynamic Database
D. Document Database
Answer(s): B Explanation:
DynamoDB is a fast, fully managed NoSQL database service. Reference:
http://aws.amazon.com/dynamodb/
QUESTION: 34
In relation to Amazon SQS, how many queues and messages can you have per queue
for each
user?
A. Unlimited
B. 10
C. 256
D. 500
Answer(s): A Explanation:
Amazon SQS supports an unlimited number of queues and unlimited number of
messages per
queue for each user. Please be aware that Amazon SQS automatical y deletes
messages that
have been in the queue for more than 4 days. Reference: https://aws.amazon.com/items/1343?externalID=1343
QUESTION: 35
Doug has created a VPC with CIDR 10.201.0.0/16 in his AWS account. In this VPC
he has
created a public subnet with CIDR block 10.201.31.0/24. While launching a new
EC2 from the
console, he is not able to assign the private IP address 10.201.31.6 to this
instance. Which is
the most likely reason for this issue?
A. Private IP address 10.201.31.6 is not part of the associated subnet's IP
address range.
B. Private IP address 10.201.31.6 is blocked via ACLs in Amazon infrastructure
as a part of
platform security.
C. Private address IP 10.201.31.6 is currently assigned to another interface.
D. Private IP address 10.201.31.6 is reserved by Amazon for IP networking
purposes.
Answer(s): C Explanation:
In Amazon VPC, you can assign any Private IP address to your instance as long
as it is:
Part of the associated subnet's IP address range
Not reserved by Amazon for IP networking purposes
Not currently assigned to another interface Reference: http://aws.amazon.com/vpc/faqs/
QUESTION: 36
Regarding Amazon SQS, are there restrictions on the names of Amazon SQS queues?
A. No
B. Yes. Queue names must be unique within an AWS account and you cannot use
hyphens (-)
and underscores (_)
C. Yes. Queue names are limited to 80 characters and queue names must be unique
within an
AWS account
D. Yes. Queue names are limited to 80 characters but queue names do not need to
be unique
within an AWS account
Answer(s): C Explanation:
Queue names are limited to 80 characters. Alphanumeric characters plus hyphens
(-) and
underscores (_) are allowed. Queue names must be unique within an AWS account.
After you
delete a queue, you can reuse the queue name. Reference: https://aws.amazon.com/sqs/faqs/
QUESTION: 37
In Amazon SNS, to send push notifications to mobile devices using Amazon SNS
and ADM, you
need to obtain the following, except:
A. Client secret
B. Client ID
C. Device token
D. Registration ID
Answer(s): C Explanation:
To send push notifications to mobile devices using Amazon SNS and ADM, you need
to obtain
the following: Registration ID and Client secret. Reference: http://docs.aws.amazon.com/sns/latest/dg/SNSMobilePushPrereq.html
QUESTION: 38
Regarding Amazon SNS, to begin using Amazon SNS mobile push notifications, you
first need
__________that uses one of the supported push notification services: APNS, GCM,
or ADM.
A. an access policy for the mobile endpoints
B. to active push notification service of Amazon SNS
C. to know the type of mobile device operating system
D. an app for the mobile endpoints
Answer(s): D Explanation:
In Amazon SNS, to begin using Amazon SNS mobile push notifications, you first
need an app
for the mobile endpoints that uses one of the supported push notification
services: APNS, GCM,
or ADM. After you've registered and configured the app to use one of these
services, you
configure Amazon SNS to send push notifications to the mobile endpoints. Reference: http://docs.aws.amazon.com/sns/latest/dg/SNSMobilePush.html
QUESTION: 39
How many types of block devices does Amazon EC2 support?
A. 5
B. 1
C. 2
D. 4
Answer(s): C Explanation:
Amazon EC2 supports 2 types of block devices. Reference:
http://docs.amazonwebservices.com/AWSEC2/latest/UserGuide/block-device-mapping-
concepts.html
QUESTION: 40
ExamKiller (with AWS account ID 111122223333) has created 50 IAM users for its
organization's employees. ExamKil er wants to make the AWS console login URL
for all IAM
users as: https://
examkil er.signin.aws.amazon.com/console/. How can this be configured?
A. Create a bucket with the name ExamKil er and map it with the IAM alias
B. It is not possible to have capital letters as a part of the alias name
C. The user needs to use Route 53 to map
the ExamKil er domain and IAM URL
D. For the AWS account, create an alias ExamKil er for the IAM login
Answer(s): B Explanation:
If a user wants the URL of the AWS IAM sign-in page to have the company name
instead of the
AWS account ID, he can create an alias for his AWS account ID. The alias must
be unique
across all Amazon Webservices products and contain only digits, lowercase letters,
and
hyphens. Reference: http://docs.aws.amazon.com/IAM/latest/UserGuide/AccountAlias.html
QUESTION: 41
Can a user get a notification of each instance start / terminate configured
with Auto Scaling?
A. Yes, always
B. No
C. Yes, if configured with the Auto Scaling group
D. Yes, if configured with the Launch Config
Answer(s): C Explanation:
The user can get notifications using SNS if he has configured the notifications
while creating the
Auto Scaling group. Reference:
http://docs.aws.amazon.com/AutoScaling/latest/DeveloperGuide/GettingStartedTutorial.html QUESTION: 42
AutoScaling is configured with 3 AZs. Each zone has 5 instances running. If
AutoScaling wants
to terminate an instance based on the policy action, which instance wil it
terminate first?
A. Terminate the first launched instance
B. Randomly select the instance for termination
C. Terminate the instance from the AZ which does not have a high AWS load
D. Terminate the instance from the AZ which has instances running near to the
bil ing hour
Answer(s): B Explanation:
Before Auto Scaling selects an instance to terminate, it first identifies the
Availability Zone that
has more instances than the other Availability Zones used by the group. If all
the Availability
Zones have the same number of instances, it identifies a random Availability
Zone. Reference: http://docs.aws.amazon.com/AutoScaling/latest/DeveloperGuide/us-termination-policy.html
QUESTION: 43
In regard to DynamoDB, can I delete local secondary indexes?
A. Yes, if it is a primary hash key index
B. No
C. Yes, if it is a local secondary indexes
D. Yes, if it is a Global secondary indexes
Answer(s): B Explanation:
In DynamoDB, an index cannot be modified once it is created. Reference: http://aws.amazon.com/dynamodb/faqs/#security_anchor
QUESTION: 44
You need to develop and run some new applications on AWS and you know that
Elastic
Beanstalk and CloudFormation can both help as a deployment mechanism for a
broad range of
AWS resources. Which of the following statements best describes the differences
between
Elastic Beanstalk and CloudFormation?
A. Elastic Beanstalk uses Elastic load balancing and CloudFormation doesn't.
B. CloudFormation is faster in deploying applications than Elastic Beanstalk.
C. CloudFormation is much more powerful than Elastic Beanstalk, because you can
actual y
design and script custom resources
D. Elastic Beanstalk is faster in deploying applications than CloudFormation.
Answer(s): C Explanation:
These services are designed to complement each other. AWS Elastic Beanstalk
provides an
environment to easily develop and run applications in the cloud. It is
integrated with developer
tools and provides a one-stop experience for you to manage the lifecycle of
your applications.
AWS CloudFormation is a convenient deployment mechanism for a broad range of
AWS
resources. It supports the infrastructure needs of many different types of
applications such as
existing enterprise applications, legacy applications, applications built using
a variety of AWS
resources and container-based solutions (including those built using AWS
Elastic Beanstalk).
AWS CloudFormation introduces two new concepts: The template, a JSON-format,
text-based
file that describes all the AWS resources you need to deploy to run your
application and the
stack, the set of AWS resources that are created and managed as a single unit
when AWS
CloudFormation instantiates a template. Reference: http://aws.amazon.com/cloudformation/faqs/
QUESTION: 45
Can you SSH to your private machines that reside in a VPC from outside without
elastic IP?
A. Yes, but only if you have direct connect or vpn
B. Only if you are using a non-US region
C. Only if you are using a US region
D. No
Answer(s): A Explanation:
The instances that reside in the private subnets of your VPC are not reachable
from the Internet, meaning that is not possible to ssh into them. To
interact with them you can use a
bastion server, located in a public subnet, that will act as a proxy for
them.
You can also connect if you have direct connect or vpn. Reference: http://docs.aws.amazon.com/AmazonVPC/latest/UserGuide/VPC_Scenario2.html
QUESTION: 46
Does AWS CloudFormation support Amazon EC2 tagging?
A. It depends if the Amazon EC2 tagging has been defined in the template.
B. No, it doesn't support Amazon EC2 tagging.
C. No, CloudFormation doesn't support any tagging
D. Yes, AWS CloudFormation supports Amazon EC2 tagging
Answer(s): D Explanation:
In AWS CloudFormation, Amazon EC2 resources that support the tagging feature
can also be
tagged in an AWS template. The tag values can refer to template parameters,
other resource
names, resource attribute values (e.g. addresses), or values computed by simple
functions
(e.g., a concatenated list of strings). Reference: http://aws.amazon.com/cloudformation/faqs/
QUESTION: 47
A user has created a MySQL RDS instance. Which of the below mentioned options
is
mandatory to configure while creating an instance?
A. Multi AZ deployment setup
B. Automated backup window
C. Availability Zone
D. Maintenance window
Answer(s): A Explanation:
When creating an RDS instance, the user needs to specify whether it is Multi AZ
or not. If the
user does not provide the value for the zone, the maintenance window or
automated backup
window, RDS wil automatical y select the value. Reference: http://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/Concepts.MultiAZ.html
QUESTION: 48
A user has enabled the automated backup, but not specified the backup window.
What wil RDS
do in this case?
A. Wil throw an error on instance launch
B. RDS wil take 3 AM - 3:30 AM as the default window
C. RDS assigns a random time period based on the region
D. Wil not allow to launch a DB instance
Answer(s): C Explanation:
If the user does not specify a preferred backup window while enabling an
automated backup,
Amazon RDS assigns a default 30-minute backup window which is selected at
random from an
8-hour block of time per region. Reference:
http://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/Overview.BackingUpAndRestoring
AmazonR DSInstances.html
QUESTION: 49
A user is planning to host a web server as well as an app server on a single
EC2 instance which
is a part of the public subnet of a VPC. How can the user setup to have two
separate public IPs
and separate security groups for both the application as well as the web
server?
A. Launch a VPC instance with two network interfaces. Assign a separate
security group to
each and AWS wil assign a separate public IP to them.
B. Launch VPC with two separate subnets and make the instance a part of both
the subnets.
C. Launch a VPC instance with two network interfaces. Assign a separate
security group and
elastic IP to them.
D. Launch a VPC with ELB such that it redirects requests to separate VPC
instances of the
public subnet.
Answer(s): C Explanation:
If you need to host multiple websites(with different IPs) on a single EC2
instance, the following
is the suggested method from AWS.
Launch a VPC instance with two network interfaces
Assign elastic IPs from VPC EIP pool to those interfaces (Because, when the
user has attached
more than one network interface with an instance, AWS cannot assign public IPs
to them.)
Assign separate Security Groups if separate Security Groups are needed This
scenario also
helps for operating network appliances, such as firewalls or load balancers
that have multiple
private IP addresses for each network interface. Reference: http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/MultipleIP.html
QUESTION: 50
An online gaming site asked you if you can deploy a database that is a fast,
highly scalable
NoSQL database service in AWS for a new site that he wants to build. Which
database should
you recommend?
A. Amazon Redshift
B. Amazon SimpleDB
C. Amazon DynamoDB
D. Amazon RDS
Answer(s): C Explanation:
Amazon DynamoDB is ideal for database applications that require very low
latency and predictable performance at any scale but don't need complex
querying capabilities like joins or
transactions. Amazon DynamoDB is a fully-managed NoSQL database service that
offers high
performance, predictable throughput and low cost. It is easy to set up,
operate, and scale. With
Amazon DynamoDB, you can start small, specify the throughput and storage you
need, and
easily scale your capacity requirements on the fly. Amazon DynamoDB
automatically partitions
data over a number of servers to meet your request capacity. In addition,
DynamoDB
automatically replicates your data synchronously across multiple Availability
Zones within an
AWS Region to ensure high-availability and data durability. Reference: https://aws.amazon.com/running_databases/#dynamodb_anchor
QUESTION: 51
How long are the messages kept on an SQS queue by default?
A. If a message is not read, it is never deleted
B. 2 weeks
C. 1 day
D. 4 days
Answer(s): D Explanation:
The SQS message retention period is configurable and can be set anywhere from 1
minute to 2
weeks. The default is 4 days and once the message retention limit is reached
your messages
wil be automatically deleted. The option for longer message retention provides
greater flexibility
to allow for longer intervals between message production and consumption. Reference: https://aws.amazon.com/sqs/faqs/
QUESTION: 52
Regarding Amazon SWF, the coordination logic in a workflow is contained in a
software
program called a ________.
A. Handler
B. Decider
C. Cordinator
D. Worker
Answer(s): B Explanation:
In Amazon SWF, the coordination logic in a workflow is contained in a software
program called
a decider. The decider schedules activity tasks, provides input data to the
activity workers,
processes events that arrive while the workflow is in progress, and ultimately
ends (or closes)
the workflow when the objective has been completed. Reference: http://docs.aws.amazon.com/amazonswf/latest/developerguide/swf-dg-intro-to-swf.html
QUESTION: 53
A user has attached one RDS security group with 5 RDS instances. The user has
changed the ingress rule for the security group. What wil be the initial
status of the ingress rule?
A. Approving
B. Implementing
C. Authorizing
D. It is not possible to assign a single group to multiple DB instances
Answer(s): C Explanation:
When the user makes any changes to the RDS security group the rule status will
be authorizing
for some time until the changes are applied to all instances that the group is
connected with.
Once the changes are propagated the rule status wil change to authorized. Reference:
http://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/USER_WorkingWithSecurityGroup
s.html
QUESTION: 54
A user has attached an EBS volume to a running Linux instance as a
"/dev/sdf" device. The
user is unable to see the attached device when he runs the command "df
-h". What is the
possible reason for this?
A. The volume is not in the same AZ of the instance
B. The volume is not formatted
C. The volume is not attached as a root device
D. The volume is not mounted
Answer(s): D Explanation:
When a user creates an EBS volume and attaches it as a device, it is required
to mount the
device. If the device/volume is not mounted it wil not be available in the
listing. Reference: http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/AmazonEBS.html
QUESTION: 55
A user has setup an application on EC2 which uses the IAM user access key and
secret access
key to make secure calls to S3. The user wants to temporarily stop the access
to S3 for that
IAM user. What should the root owner do?
A. Delete the IAM user
B. Change the access key and secret access key for the users
C. Disable the access keys for the IAM user
D. Stop the instance
Answer(s): C Explanation:
If the user wants to temporarily stop the access to S3 the best solution is to
disable the keys.
Deleting the user wil result in a loss of all the credentials and the app will
not be useful in the
future. If the user stops the instance IAM users can stil access S3. The change
of the key does
not help either as they are stil active. The best possible solution is to
disable the keys.
