0

Scalability in the data plane is easy to achieve by simply having more replicas. The configuration the proxies receive from the control plane makes that possible . However, if the number of services (and their sidecar proxies) get too big, the control plane can become a bottleneck by not being able to send metrics or other configuration data fast enough to the sidecars.

Is there a possibility to horizontally scale the control plane of a service mesh so that the total amount of possible services can be increased? I didn't find anything regarding that.

1

It's true that when you scale your services is possible to have a bottleneck on the monitoring, but for avoiding that you have to check if your monitoring system is pull or push, in case it is push maybe is not the best option but pull is not the right answer for all situations.

One of the most used monitoring systems today is prometheus, you can se how it works here:

https://prometheus.io/docs/introduction/overview/

But is not the only one and you need to see how it should fit in your stack.

Check the possible integrations with istio

https://istio.io/docs/tasks/telemetry/metrics/querying-metrics/

In case you have already some of that probably need to think on scaling this part instead of the sidecars or istio itself.

1
  • I know that you can scale your monitoring in Istio, but how do you scale e.g. Pilot? It is not a strictly practical question. I ask to know the limits of a service mesh. It could be related to a specific one, like Istio, though.
    – jhamfler
    Jul 24 '19 at 9:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.