[GLBC] Max RPS should be per instance instead of per instance group

[nicksardo]

Without knowing how much traffic a k8s node can handle, the best strategy is to spray traffic across all nodes. The GLBC currently sets MaxRPS to 1 per instance group; however, this is not ideal because in multi-zonal clusters IGs could be larger than others.

We should also question whether 1 is an ideal RPS. Is there any LB optimization to increasing the value but still remaining small?

@thockin

[thockin]

I think one per instance is a good first step. Beyond that, we might need to get user-input on balancing policy decisions like this, which falls somewhat outside core Ingress for now.

…

On Tue, Mar 21, 2017 at 11:51 AM, Nick Sardo ***@***.***> wrote: Without knowing how much traffic a k8s node can handle, the best strategy is to spray traffic across all nodes. The GLBC currently sets MaxRPS to 1 per instance group; however, this is not ideal because in multi-zonal clusters, some instance groups could be larger than others. We should also question whether 1 is an ideal RPS. Is there any LB optimization to increasing the value but still remaining small? @thockin <https://github.com/thockin> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#475>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AFVgVAAfO6XxUJHbb7N0cpkIegnGDYkjks5roBxPgaJpZM4MkQJI> .

[nicksardo]

Sounds good.

I'll throw you a PR for this after backside re-encryption goes in.