Istio and big data


#1

Hi,

I’m wondering if there is performance penalty by using istio. The proxy is running in sidecar container. Is it suitable for big data scenario. We have recording service (take incoming video stream and save to disk), as well as services to continuously listen and process large amount of IOT events.

How does Istio affect network throughput and latency?

Thanks,
Tao


#2

In our testing it is dependent on the kind of services you are running. It improves the latency if all the services are within the same cluster and sending traffic to each other. The latency degrades if your main traffic is going in and out of cluster. Istio is a service mesh, it is designed to take advantage of thousands of micro services within the same cluster.


#3

Thank you very much for the information. How about network throughput, does it affect network throughput?


#4

I would point you also Istio performance benchmark to have an idea: https://istio.io/docs/concepts/performance-and-scalability/

Performance depends mainly on your application architecture and deployment scenario (number of proxies involved end to end, tracing, logging …). For such reason, envoy doesn’t publish performance benchmark result.

Please note also that if you are handling microservices concerns (circuit breaker, retry strategy traffic splitting, authentication, tracing … ) on the code level, you will for sure compensate the eventual overhead introduced by Istio by migrating these capabilities to be handled declaratively by Istio.