You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Status cache is by far the biggest bottleneck right now if you ingest many txs regardless of CUs, because it's an early bottleneck
It's using a rwlock now. The short term fix is to use a concurrent hash table. Dashmap would work, but Alessandro has been benchmarking https://docs.rs/flurry/latest/flurry and it's much much faster.
However, the existing design doesn’t scale well to 1M TPS without pre-allocating a ton of memory.
Status cache is by far the biggest bottleneck right now if you ingest many txs regardless of CUs, because it's an early bottleneck
It's using a rwlock now. The short term fix is to use a concurrent hash table. Dashmap would work, but Alessandro has been benchmarking https://docs.rs/flurry/latest/flurry and it's much much faster.
However, the existing design doesn’t scale well to 1M TPS without pre-allocating a ton of memory.
The text was updated successfully, but these errors were encountered: