Brennan

Activated on mainnet at ~2024/03/11/2:28PM PT The cluster was spending ~14% of the time hashing before and now is spending ~20% of the time, so increasing linearly with hash requirement...

> can i work on this feature? The feature is effectively done. Plan is to enable these features (for which I hold the key) to slowly increase required number of...

I think we can close this issue now that `Bank.rs` is well under the 20k syntax highlighting threshold. Any objections?

+1 to what @behzadnouri said. From my observations, the vast majority (90%+) of gossip vote transactions error out for `already_processed` because generally turbine votes land faster. However, in the forking...

I'm observing on the order of 100 shred repair requests being issued per second, but the vast majority of these end up not getting inserted into the blockstore because they...

A few more interesting observations: 1. I'm seeing that a small amount of repairs completing successfully are requested via the `HighestShred` repair type through the `generate_repairs_for_slot` function. These repairs seem...

2 actions that can be taken to cut down on repair related network traffic: 1. Increase `MAX_TURBINE_PROPAGATION_IN_MS` so that we wait for longer for shreds to show up through turbine...

> Are the stats gathered in [#28638 (comment)](https://github.com/solana-labs/solana/issues/28638#issuecomment-1362099338) with current 100ms `MAX_TURBINE_PROPAGATION_IN_MS` or some other setup? They are gathered with `MAX_TURBINE_PROPAGATION_IN_MS` set to 200. However, I also have stats gathered...

> This is another idea we can try. > > 1. On a given node (repairer), monitor how many repaired shreds are already in the blockstore by the time repaired...

> I added some counters to see how many duplicate shreds are because of repair. Running on monogon, with 100 nodes, 1 client, and ` --bench-tps-args 'tx-count=15000 threads=4 thread-batch-sleep-ms=0'` produces...