Download to read offline
Uploaded byOkis Chuang
Let's Talk About Serverless - Focusing on AWS Lambda
This document provides an overview of serverless computing using AWS Lambda. It discusses key concepts such as Lambda functions being stateless and event-triggered. Examples are given of using Lambda for various use cases including user registration with DynamoDB and API Gateway. Limitations around resources and retries are also covered. The document concludes with pros and cons of the serverless model and an example of how it could be applied to a security product.
More Related Content
Similar to Let's Talk About Serverless - Focusing on AWS Lambda
Let's Talk About Serverless - Focusing on AWS Lambda
- 1.
- 2. No server iseasier to manage than no server. Werner Vogels
- 3.
- 4.
- 5. 什麼是 Serverless ?什麼是 Lambda ? ▸Serverless can also mean applications where some amount of server-side logic is still written by the application developer but unlike traditional architectures is run in stateless compute containers that are event-triggered, ephemeral (may only last for one invocation) ▸AWS Lambda is a compute service that lets you run code without provisioning or managing servers http://martinfowler.com/articles/serverless.html
- 6.
- 10. Event handler 內的兩個參數 ▸Event ▸Triggerlambda 的 event 來源 ▸s3, sns, API gateway, DynamoDB…etc. ▸Context ▸跟 Lambda service 溝通可拿到的 runtime information ▸AWS 停掉 function 前剩餘的時間 ▸Cloudwatch log group and log stream ▸觸發 function 執行的 requestID (可用來查詢 Support) ▸AWS mobile SDK 可以拿到 mobile 的資訊
- 11.
- 12. 文字 Lambda 的限制 1/3 ▸必須設計成stateless ▸有最大memory限制 ▸有最大執行時間限制
- 13.
- 14.
- 15. Lambda 到底是怎麼搞的？ ▸每個 function在獨立 container sandbox 中執行，resource, configuration 與其他 function 分開 ▸Lambda Container Reuse (freeze-thaw cycle) ▸Lambda Host 在運行 Function 時，會重覆利用 Container ；Container 若正常結束，Lambda 會保存當下狀態 (freeze) 備日後復用 ▸Function 再次執行時，若該 Frozen Container 仍然可用， Lambda 會重新啟動 (thaw) 並執行 handler，無需再初始化 https://aws.amazon.com/tw/blogs/compute/container-reuse-in-lambda/
- 16. Lambda 到底是怎麼搞的？ ▸對岸神人自幹 Lambda ▸LambdaDocker ▸Docker + Swarm ▸event-driven code runtime like AWS Lambda service ▸https://github.com/tobegit3hub/lambda-docker
- 17. Retries on Errors ▸Lambdafail 的三種原因 ▸Times out ▸Fails to parse input data ▸Reaches resource constraints: memory, other timeout ▸Retry ▸Non-stream-based event source ▸SYNC: invoking application receive 429 error ▸ASYNC: DLQ (Dead Letter Queue) 沒設定 event 會被 discard ▸Stream-based even source ▸Kinesis, DynamoDB ▸lambda 會持續 retry 拉資料直到 data expired，為了確保 order 會是 blocking 狀態 (不會繼續讀新資料直到 failed batch 成功做完或 expired)
- 18. Avoid Chaos ▸Lambda versioncontrol ▸stage variable ▸良好且固定的 naming rule ▸加上 prefix，參考 arn 命名規則 ▸使用 deployment tool ▸aws cli ▸cloudformation ▸serverless framework: serverless, Apex…etc ▸Orchestration生態尚未成熟，需暫時自己手刻 API Gateway Lambda FQDN /prod/Resources auth:stable https://api.diamond /dev/Resources auth:$LATEST https://api- dev.diamond /qa/Resources auth:qa https://api- qa.diamond
- 19.
- 20.
- 21. 各種 Use case ▸使用者註冊: API GW + Lambda + DynamoDB + SES ▸使用者認證 : API GW + Lambda + Cognito/IAM ▸RESTful API : API GW + Lambda ▸各種 event-driven 使用 lambda ▸Static website hosting : Cloudfront + S3 + Lambda ▸Log parsing/monitoring : Cloudwatch + Lambda ▸Data Analyzing : Kinesis + Lambda
- 22. DEMO 03-Rate-Based blacklisting with AWSWAF and Lambda
- 23. SCENARIO Without AWS WAF Webserver User Attacker
- 24. SCENARIO With AWS WAF Webserver User Attacker
- 26.
- 27. Pros & Cons ▸Costcould be reduced ▸Demand based 的計價方式更適合 bursty traffic 類型 ▸Scale 更容易 ▸更容易 maintain 每個 component ▸Cost could be increased ▸在 design 的時候就必須有拆分小 service 的 overhead ▸Local development and unit testing 不容易 ▸realtime end-to-end testing ▸docker-lambda (很像 AWS lambda 環境的 docker image, test runners)
- 28. 文字 以 Nebula 舉個例子 ▸可以將task 拆分為兩大類 ▸event-driven: security event, user-triggered event, scan result reporting…etc. ▸periodical event: profile pulling, HW health reporting, daily scan reporting…etc. ▸拆開以後 (decoupled) ▸easier to deploy ▸easier to try new things ▸降低 long-run EC2 instance 的規格
- 29.
- 30.
Editor's Notes
- #4 3-tier v.s. micro-services (serverless)
- #13 这些限制要求你把每个lambda函数设计得尽可能简单，一次只做一件事，但把它做到最好。很符合unix的哲学。反过来，这些限制强迫你接受极简主义之外，为你带来了无限扩容的好处。
- #14 这些限制要求你把每个lambda函数设计得尽可能简单，一次只做一件事，但把它做到最好。很符合unix的哲学。反过来，这些限制强迫你接受极简主义之外，为你带来了无限扩容的好处。
- #15 这些限制要求你把每个lambda函数设计得尽可能简单，一次只做一件事，但把它做到最好。很符合unix的哲学。反过来，这些限制强迫你接受极简主义之外，为你带来了无限扩容的好处。
- #18 Therefore, if a Lambda function fails, AWS Lambda attempts to process the erring batch of records until the time the data expires, which can be up to seven days for Amazon Kinesis Streams.
- #23 idea is raised from diamond au package cdn