Best of Both Worlds: Exploring the Role of Centralisation in IPFS
summary
The system introduces three semi-central components: IPNI indexers that index ~100× more provider records than the DHT and publish advertisements containing ~10^4 CIDs each, indexing a provider record in ~500 µs versus >10 s in the DHT; Hydra boosters that multiplex thousands of virtual heads into a shared provider DB to cut DHT lookup latency (shutting Hydra causes a measurable spike and revealed ~60% of nodes were unreachable); and HTTP gateways (≈2.8M clients/day, 120M requests) handling 32% of traffic and 90% of requests from mobile.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:12Decentralizing IPFS challenges
- 4:40Content publication bottleneck
- 5:43Content retrieval performance
- 6:25Mobile adoption hurdles
- 7:09Hybrid components
- 11:29Hybrid solution performance
- 16:51Trade-offs of hybrid approaches
- 19:01Hydra boosters success
- 20:46Conclusion
- 22:10Q&A
related talks
27mDecentralized Storage with IPFS: How does it work under the hood?
07mState of IPFS Public Utilities
16mIPFS and Dapps: Obstacles and Opportunities
30m🛠️ IPFS in JS; Helia ramp-up
30m🛠️ IPFS in JS; Helia ramp-up
01h 00mIPFS — InterPlanetary File System with Node.js
18mIPIP Protocol Improvements: Review & Working Session
12mThe State of IPFS in JS