On January 19, 2024 we used our global RPC Performance Inspector app to compare latencies for 17 public endpoints on Optimism Mainnet from 27 locations on 6 continents. Our goal was to identify the best Optimism Mainnet public RPC endpoints.
Today we launched public infrastructure performance analytics for the Top 11 Optimism public RPC providers from our earlier analysis: 1RPC, Allnodes, Ankr, Blast API, BlockPI, Gateway.fm, LlamaNodes, Nodies, Optimism Foundation, Tenderly and thirdweb.
Methodology
Both the Performance Inspector app and the newly launched Performance Public analytics page use our remote RPC probes in 27 AWS regions. The probes make a warmup RPC request which is omitted from the results. Its purpose is to exclude the time for DNS lookup and SSL negotiation.
Performance Inspector vs. Public Infrastructure Analytics
The key differences between the two performance tools can be summarized as:
- Anyone can run the inspector on-demand for 30 seconds. Continental and global results are arithmetic means (averages) of the underlying AWS regions.
- The public performance analytics are collected perpetually, in 1-minute intervals for a group of endpoints selected to represent the ecosystem. Its line charts are medians of the underlying results. The bar chart rankings are averages of the respective line chart over the corresponding time horizon (e.g. 30 minutes).
The global chart & ranking requires each endpoint to have at least one success in each continental group, in each time slice. Otherwise, the endpoint with excessive fails is omitted from the current global ranking. One such example is illustrated in these screenshots (taken around 19:50:52 UTC on Sat, 03 Feb 2024). Please note that LLamaNodes is No. 10 in Asia (median of 1-8 regions):
But, LLamaNodes is not in the Global ranking for the same time horizon:
…because LLamaNodes had errors in Australia during the slice “:32”. You can see their yellow line is broken to the left of the marker:
Lastly, the LLamaNodes errors are implied in the screenshot of the details table. Its purpose is to show regional measures for the most recent slice of time. Observe the superscript “12” which is the total count of successful measures from the possible 27:
Additional information about each computation is available in each FAQ.
Our roadmap is full of improvements. Soon we will show counts of RPC errors as well as longer horizons including 24hours, 10days and 30days.
