Due to the way merge-split compaction works, each compactor instance can put significant pressure on the network, especially if the number of compaction shards is high. To mitigate this, we should implement traffic shaping. We already have an option to limit the concurrency (compactor.block-sync-concurrency, which is set to 8 by default), but this isn't helpful – regardless of the number of concurrent threads, 100% of the network is utilized
Due to the way merge-split compaction works, each compactor instance can put significant pressure on the network, especially if the number of compaction shards is high. To mitigate this, we should implement traffic shaping. We already have an option to limit the concurrency (
compactor.block-sync-concurrency, which is set to 8 by default), but this isn't helpful – regardless of the number of concurrent threads, 100% of the network is utilized