Andrei Sandu
Andrei Sandu
This seems to be a tractable problem as per the retreat discussions, but there might be dragons and a huge amount of work, but it would be really worth it...
Alternatively we could be doing his with cgroups and would also work in native mode.
We can use `describe pod` to detect when the pod scheduler is waiting for more capacity in the cluster.
Currently it is hard to reason about how the test is going in terms of latency. It would be useful to have each line of output be prefixed with the...
We're looking at writing an integration test suite that focuses on performance testing, more specifically on a list of key indicators that are covered in https://github.com/paritytech/polkadot-sdk/issues/874. The current design of...
It would also make sense to define a probability density function to make it look more chaotic/realistic.
This will enable us determine the current performance characteristics of nodes and test networks and prevent performance regressions. The goal here is to ensure the following expectations are met for...
Previously you would need to spend some time going into the code to figure out if a task is blocking or async. This change allows this determination directly from metrics.
Currently we are not preferring the fast path and this starts to be a problem above 300 para validators. PoV recovery times distribution screenshot below. We need to dive deeper...
fixes paritytech/polkadot-sdk#729 This PR will upgrade the network protocol to version 2 -> `VStaging` which will later be renamed to `V2`. This version introduces a new kind of assignment certificate...