Redis-OM Lambda CDK Sample Application

This sample shows how to define a Lambda function which uses Redis-om via AWS Cloud Development Kit. This is a fork of prisma-lambda-cdk sample repo which replaces mysql/postgres with redis stack & prisma with Redis-om.

Redis OM is an open source object mapping, search etc library for Redis written in Typescript & very useful with its developer friendly API.

By this sample, you can see how a Lambda function which uses Redis OM can be deployed with CDK, how indexes can be created from Lambda, and how Redis OM works in Lambda.

Lambda Dashboard

How it works

Architecture

This sample consists of the following AWS services:

• Amazon VPC
• Redis Cloud (Amazon doesn't offer redis stack yet)
• AWS Lambda

There're two Lambda functions:

1. a function that reads and writes records from/to Redis handler.ts
2. a function that sets up the index for Request entity migration-runner.ts

How the data is stored:

Database Schema

class Request extends Entity {
  public awsRequestId: string;
  public createdAt: Date;
}

The Request data is stored in JSON format using RedisJSON component where each entry contains the following values:

• awsRequestId: unique request id generated by aws
• createdAt: Timestamp when this request was received

It uses the repository.createAndSave() function which internally uses JSON.SET command. Code Example:

requestRepository.createAndSave({
  awsRequestId: context.awsRequestId,
  createdAt: new Date(),
});

How the data is accessed:

• The request data is accessed via Request:{entityId} key. RedisSearch module is used to fetch the data from Redis.

It uses the repository.search.all() function which internally uses FT.SEARCH command. Code Example:

requestRepository.search().all();

Performance Benchmarks

Note: All the benchmarking was done using k6

Terminology:

• VU Count: Number of concurrent requests being made
• Duration: The duration for which test was run
• Function type: Type of Lambda function executed
• Prisma p95: p95 response time of prisma lambda function
• Redis OM p95: p95 response time of redis om lambda function
• Prisma p90: p90 response time of prisma lambda function
• Redis OM 90: p90 response time of redis om lambda function
• Prisma AVG: Avg response time of prisma lambda function
• Redis OM AVG: Avg response time of redis om lambda function

VU Count	Duration	Function Type	Prisma p95	Redis OM P95	Prisma p90	Redis OM p90	Prisma AVG	Redis OM AVG	Prisma Failed Reqs	Redis OM Failed Reqs
10	30s	Handler - Create Data	91.99 ms	67.56 ms	74.57 ms	60.4 ms	71.56 ms	57.16 ms	0	0
20	30s	Handler - Create Data	89.35ms	65.7ms	74ms	59.19ms	70.56ms	54.73ms	26.60%	24.66%
20	30s	Handler - Retrieve Data	698.56ms	516.26ms	600.07ms	394.06ms	547.38ms	344.9ms	0	0
20	30s	Migration Runner	10.87s	3.16s	8.95s	2.93s	7.8s	2.67s	0	0

From these stats we can infer that there's a signifact performance boost when migrating schemas using Redis OM instead of Prisma. Reads from Redis are also faster than Mysql & while not much but writes are faster as well.

How to run it locally?

Prerequisites

• NodeJs - V16
• Docker
• AWS SDK

Local installation

• Setup aws CDK by following this guide
• copy .env.example & add the REDIS_URI to it

cp .env.example .env

• Install dependencies using npm

npm ci

• Deploy the Lambda function using cdk

npx cdk deploy --require-approval never

• Take note of the output generated by the above function so that we can use them to test the lambda functions

1. RedisOmLambdaCdkStack.ApplicationHandlerLambdaArn*
2. RedisOmLambdaCdkStack.ApplicationMigrationRunnerLambdaArn*

• Create the search index

aws lambda invoke --function-name ApplicationMigrationRunnerLambdaArn response.json
# replace the `ApplicationMigrationRunnerLambdaArn` with actual ARN you wrote down in the previous step

• Creating Records

aws lambda invoke --function-name ApplicationHandlerLambdaArn response.json
# Replace the `ApplicationHandlerLambdaArn` with actual ARN you wrote down in the previous step

• Fetching Records

aws lambda invoke --function-name ApplicationHandlerLambdaArn --payload eyJjb21tYW5kIjoiZ2V0In0= response.json
# Replace the `ApplicationHandlerLambdaArn` with actual ARN you wrote down in the previous step
# Check response.json to see the records you created