getHeader and submitBlindBlock latencies

[terencechain]

Not sure if this is the best place to track this. Feel free to close this and let me know a more appropriate place, we can move the issue somewhere else.

I've been observing high latencies with both getStatus, getHeader and submitBlindBlock calls using the Kiln relay.

getStatus, between 100-500ms

getHeader, this gotten worse in the last 24 hours, many calls exceeded 1-2s

submitBlindBlock, mostly around 200ms

This is a prysm beacon node getting a beacon block for validator to sign. We are seeing 10-25x latency increases

I understand that many moving parts are currently under optimized, but I still want to open an issue for tracking

[metachris]

That's great, thank you for reporting. Well investigate today and update here.

[metachris]

We've updated our Relay and builder infrastructure across all the supported testnets (Kiln, Ropsten and Sepolia), and we're seeing very quick response timings now (in the single-digit milliseconds mostly). Also it's not necessary anymore to batch validator registrations in small chunks, you can many registrations with a single call without problems.

Let us know if you can confirm the numbers from your end, or if we need to further investigate network delays.

[terencechain]

getHeader has gotten much better 🎉

submitBlindBlock and getStatus are pretty much the same as before. 95% percentile still around ~200-500ms.
We won't be calling getStatus anymore. Only once during startup so that's not going to be an issue for us. Let me know if you want me to close this and just track submitBlindBlock separately

[tbenr]

Last 24h for us on Ropsten.
Focus on the pies, we had 6 slow responses (between 1s and 2s) for GetHeader and 3 slow (between 300 and 500 ms) for GetPayload

[tbenr]

Last 24h for us on Sepolia.

[tbenr]

registration response time improved significantly

[metachris]

@tbenr @terencechain do you have any current charts for the last days/week?

[tbenr]

Last 2 days in sepolia:

getPayload: