General Recommendations
Practical guidance for production Shreder workloads
Recommendations that apply across Shreder streaming products.
Pick the right stream
- Use Geyser gRPC / Fastlane when you need 100% compatible with standard Yellowstone Geyser gRPC and post-execution context.
- Use Decoded Shreds (ShredStream gRPC) for early pre-execution transactions without a local deshredder.
- Use Binary for compact
VersionedTransactionbytes and the lowest shred-path latency. - Add Preconfs when you need earlier-than-shred signals on top of Binary.
- Use Raw Shreds when you need UDP packets and run your own decoder.
Keep filters narrow
Subscribe only to the accounts or stream types you process. Narrow filters reduce bandwidth, CPU, and queueing delay.
Colocate with the region
Deploy clients close to the Shreder endpoint you use (Frankfurt, Amsterdam, New York, London, Tokyo). Cross-region hops often dominate observed latency.
Regional coverage
If you start with a single region, use Frankfurt — it currently carries the most transactions. We recommend at least three regions: Frankfurt, Amsterdam, and New York. When those regions perform well, add more.
Validate with your own clocks
- Raw Shreds → solana-shred-perf
- Geyser gRPC / Fastlane, Decoded Shreds, Binary → GeyserBench
Plan for reconnects
Assume disconnects happen, make sure to implement reconnection logic.
Need help?
Contact the team via shreder.xyz/contact, Discord, or Telegram.