[Proposal: XX] OnFinality Q2/Q3 2024 public RPC service support for Moonbeam, Moonriver, and Moonbase Alpha

Hi @daniel_onfinality,

Thank you for providing such detailed report!
We’re glad to see efforts in performance improvements being made and a 42k row 1-minute interval data sheet tab per region and network being provided! :clap:

Looking at the data from American regions we still see room for latency improvement, in particular with regards to the massive outliers.
It looks like the changes made caused the relatively constant 250ish ms average to go significantly below 100ms (which is what we absolutely love to see!) while introducing huge outliers way above 1,000 ms at the same time.
Also it looks like these changes were then reverted towards the end of last month.
Obviously that’s just one interpretation of the low complex data provided and nothing more, so feel free to correct us with some further insights, here.

We would love to hear more about the explicit changes made and how the outliers issue could possibly be tackled or even fully resolved.

Thanks again for your efforts and for generating such detailed reports, we really appreciate that! <3
Kind regards :sparkles:

1 Like

Hi @TreasuryCouncil

Here is the OnFinality Moonbeam SLA report for June 2024

We are pleased to report that we have met the following SLAs

  • Uptime 99.9% North America
  • Uptime 99.9% Asia Pacific
  • 99% Response Times under 1s - Asia Pacific
  • 99% Response Times under 1s - America (Moonbeam, Moonriver)

However, we missed the 99% Response Times under 1-second target in North America for Moonbase Alpha. Following recent improvements, we are confident that we will achieve all targets next month.

Here’s a summary of our observations and changes over the past several months:

First, we identified that our North Virginia Moonbeam region had irregular spikes of high latency despite being very fast on average. We hadn’t picked this up in our standard performance monitoring or with tools like compare nodes, see recent comparison here , because they use averages sampled from multiple requests

In May, we temporarily swapped Moonbeam and Moonriver to our Frankfurt region while we addressed the issue

We tested an improvement on the 5th/6th of July but it wasn’t successful

We made further optimisations to the network and hardware of our North Virginia region and it has shown excellent performance since returning Moonbeam and Moonriver on the 25th of July. Moonbase Alpha, which stayed in North Virginia, shows improvement from the 19th of July

We will deduct the Q2 SLA discounts from our Q3 submission, please let me know when we can proceed

Thanks again for the opportunity to improve our services!

Kind Regards,

Daniel

3 Likes