We're pleased to announce Jitsu v1.38 "DeLorean"! This release was mainly about performance optimizations.
Redis memory optimization#
User recognition could eat up a lot of RAM. We implemented number of measures that makes Redis RAM consumption more manageable:
- Now it's possible a dedicated Redis instance for user recognition, and configure compression, auto eviction and so on
- Jitsu can compress objects
We made a number of refactorings that significantly improved how Jitsu works under a heavy load. Here's some benchmark numbers for
m5.xlarge (4 CPUs / 16 RAM):
- Stream mode: 16k RPS
- Batch mode: 12k RPS
Improved Docker Images#
Now @jitsucom/server (if deployed standalone) can be configured with environment variables too. It's very convenient for Kubernetes based deployments.
Learn about the difference between various deployment methods here
Mixpanel destination and new destination SDK#
We added Mixpanel destination. This destination is the first one based on a new Jitsu SDK.
With new Jitsu SDK anyone can implement a destination using Typescript (as npm package) and connect it to Jitsu. At the moment, we released an alpha version of SDK, and Mixpanel destination based on the SDK. Our next steps are:
- Finalize API spec (Jan 22). We will release a comprehensive documentation and step-by-step guide for implementing destination
- Allow Jitsu users to plug in any npm packages (Feb 22). At the moment, Jitsu can work only with predefined set of destinations based on SDK. We will add an ability to reference any npm package from Self-Hosted UI
- Airbyte based sources now properly works in AWS Elastic Beanstalk environment if deployed using Dockerrun.aws.json
- A lot of minor fixes