Swiggy reduced Redis cost by 60% 👀

Good Morning! This is Pranjal from Engineering Diary.

Here's what I got for you today:

• How Swiggy reduced one of its Redis Cluster cost by 60%

• How Hubspot Frontend team enabled CI/CD to deliver value

How *Swiggy reduced Redis Cluster cost by 60%

They run 1000s of ML feature jobs daily and updates feature store. That contributes to 40% of their platform maintenance cost. To reduce the cost, they found that 50-60% of job uptime goes to writing to Redis hence this was the natural place for optimization.

They were using Redis Sequencial approach which means every request waits for ACK before sending another one. By leveraging Redis Pipelining feature, they could send requests in batches and then receive ACK's asynchronously.

This helped in following:

• 36 million operations used to take 5.3 hours now takes only 41 minutes with pipelining

• 101 million operations went down from 14.5 hours to 1.6 hours

• Also, Redis Engine's CPU utilization reduced (approx 40% from the peaks) because now it doesn't have to do frequent context switching between hardware network(kernel space) and the internal data structure (user space)

This didn't mean they could infinitely increase the batch size. They implemented buffers on the client side and found a sweet spot that gave optimal performance.

How *Hubspot Frontend team enabled CI/CD to deliver value

CI - Continuous Integration refers to the use of automated testing to validate changes to a codebaseCD - Continuous Deployment aka Continuous Delivery automatically deploys to production when the tests pass and build is successfulTogether they are knows as CI/CD Pipeline

Steps taken by the Frontend team:

• Used Typescript to check for type safety at compile time on local machines. They also use Husky pre-commit hooks to automatically review code syntax and style.

• They embrace testing practices from Testing Trophy

  • Unit tests to assess business logic

  • Integration tests to assess functional logic

  • End-to-End acceptance tests to check everything works together from user's perspective

• After changes committed to remote repo, in-house system run units and integrations tests

• When tests passes and code review is done, the branch is merged to master and Orion (Hubspot's deploy system), deploys them to shared QA environment

• The new build on QA env runs the declared end-to-end tests acceptance and upon passing the build goes live by deploying on production.

It took their team about 9 months and approximately 20 new tests suites across 4 repositories to embrace Continuous Deployment.

*Swiggy - A food delivery service in India like Uber Eats and DoorDash*Hubspot - A CRM software for businesses

That's a wrap for today. Stay thirsty & see ya soon!

If you have any suggestions or questions, I would love to hear from you.

Please share with your friends and colleagues.