Here is ProbeLab’s update on the state of the IPFS network over the past few weeks. The intention of these reports is to direct efforts to parts of the protocol that need attention. Here’s the highlights:
-
The number of Amino DHT Server nodes has stabilised at ~25k: https://probelab.io/ipfs/#chart-ipfs-servers-vs-clients-ts with the number of generally online server nodes having stabilised at ~4+k nodes, following also a declining pattern over the last few months: https://probelab.io/ipfs/topology/#chart-availability-classified-ts
- Following traditional patterns, the majority of online nodes is based in Europe, North America and Asia: https://probelab.io/ipfs/topology/#chart-availability-classified-avg
-
Kubo’s “Provide Duration” has decreased significantly with the release of v0.39.0 from more than 13s to less than 1s! This improvement came as a result of “Optimistic Provide” - an optimisation proposal that was originally proposed by the ProbeLab team and in particular @dennis-tra ! See more details and what this means in detail at: https://github.com/ipfs/kubo/releases/tag/v0.39.0 and the original ideation around it here: https://discuss.libp2p.io/t/dht-optimistic-provide/1143 (from 2021!).
- There are only around 14% of Kubo nodes on v0.39.0: https://probelab.io/ipfs/topology/#chart-agent-versions-avg. The rest are strongly encouraged to upgrade to this version to benefit from this improved performance.
-
The Amino DHT Lookup performance is stable around 0.3s-0.4s, on average among all regions: https://probelab.io/ipfs/dht/#chart-ipfs-dht-lookup-performance-ts with Europe being by far the fastest, especially compared to Asia Pacific: https://probelab.io/ipfs/dht/#chart-ipfs-dht-lookup-performance-cdf
-
The IPNI performance is mostly stable and “green” with a few exceptions where CIDs are ingested with a delay of more than 5 mins (but less than 60mins) 15% of the time: https://probelab.io/ipfs/ipni/cid.contact/#chart-ipni-uptime-bar.
-
The number of daily requests to the ipfs.io and dweb.link gateways remains relatively stable, hovering above 500M per day. ProbeLab has improved its tooling to distinguish between different service classes, i.e., instead of only Served and Blocked, we now also present the percentage of Badbits requests, which were previously grouped under Blocked. We also show the number of requests that were Redirected.
-
It’s interesting to see that the vast majority (over 99%) of Blocked requests are Badbits requests, i.e., requests for flagged content that is not served by the gateways.
-
It’s also interesting to see that the vast majority of Badbits requests (over 97%) comes from Taiwan.
-
-
The TTFB of Service Worker gateway has decreased substantially (by about 0.75s-1s) for CID requests that hit the cache of gateways from any of our probing locations: https://probelab.io/ipfs/gateways/#chart-ipfs-service-worker-gateway-ttfb-ts.