Andrei Sandu

Networking tasks CPU usage is too high

4

While running load tests on Versi (with PR https://github.com/paritytech/polkadot/pull/6810) I've observed the networking tasks eat up more than 50% of the CPU with `libp2p` tasks being top consumer. Substrate task...

approval-voting: simplify v2 assignments

5

This is a child PR of https://github.com/paritytech/polkadot/pull/6782. Removing redundant information as suggested in review discussion: https://github.com/paritytech/polkadot/pull/6782#discussion_r1117750373 Changes: - Separate transcript for v1/v2 assignments - remove `sample` and `core_indices` from v2...

approval-voting: coalesce multiple approvals in a single signed message

1

This is aimed at reducing the amount of messages and work done per relay chain block. Currently we send one signed message for each candidate approved under a block as...

`approval-voting`: implement parallel processing of signature checks.

4

I've been experimenting with using a thread pool to handle VRF signature checks which appear to be most expensive operation that we are doing in approval voting. After running some...

Async backing integration & load test suite

3

The following need to be covered with Zombienet CI tests: - [x] Network protocol upgrade deploying both master and async branch (compatibility) - [x] Runtime upgrade while running both master...

`approval-voting`: merge tranche 0 assignments

4

As pointed out in https://github.com/paritytech/polkadot/issues/6610#issuecomment-1401468099 we can save a lot of CPU and reduce gossip by doing only one RelayVrfModulo signature per authority and derive multiple assignments from the output...

Network stream events shouldn't be cloned

7

Currently, the `NetworkWorker` creates one clone per event stream listener in https://github.com/paritytech/substrate/blob/master/client/network/src/service/out_events.rs#L219 . For example, in a simple Versi test with 200 validators and 40 parachains we get around a...

av-recovery: bump small pov limit to 3MiB

1

**This has become a test PR, so should not be merged.** This should fix the slow chunk based recovery during Versi Glutton test. If it works we should merge this.

Documentation needs to include examples, known issues and limitations.