Lion - dapplion

Results 390 comments of Lion - dapplion
trafficstars

> also remove unnecessary gauge metric @tuyennhv Could you do this item in a separate PR? About the optimization, is it best to do it here or on Lodestar?

First we should quantify the cost of promises. This could lead to a lot of duplicate code and make the maintenance much more annoying. So we should have a big...

@tuyennhv Please review also all the metrics taken for each gossip packet, there are a lot!. We should at the same time: - Keep only necessary metrics: easy - Check...

@achingbrain That's a good point! Shouldn't be expensive to do

@tuyennhv The resolution of the buckets in `gossipsub_heartbeat_duration_seconds_bucket` is really poor: [0.01, 0.1, 1]. So the quantile function can show widely inaccurate results. We should improve the bucket values and...

![Screenshot from 2022-05-28 14-12-48](https://user-images.githubusercontent.com/35266934/170814886-e4310e12-f712-4af3-9b5b-09cddfd79431.png) This chart is also a bit worrying: heartbeat should not be skipped, and should run 1/0.7 = 1.42 times / sec on average. Smooth chart below...

The only hard data we have now is how many heartbeats take > 100ms to complete and how many take > 1000ms to complete. However you can see that many...

Why is emitGossip so costly? This needs benchmark testing. emitGossip is taking 2% of total CPU time ![Screenshot from 2022-06-02 12-23-10](https://user-images.githubusercontent.com/35266934/171558617-608dbaa4-7d10-4999-a9e3-69b78845cdc3.png)

@wemeetagain Is this done?

The aggregation is divided into two steps, first per committee and then a final aggregation. Maybe we could benchmark 32, 128 and 256. More than that the benchmarks would start...