Jito reports fewer small blocks on BAM validators
According to Jito, the share of small BAM blocks fell by about 77 percent, from about one block in 800 to one in 3,600. The measurement comes from the BAM Performance Team. It has not been checked independently.
On 26 October, Jito wrote on X that small blocks from BAM were down by about 77 percent. The figure is in a technical post by the BAM Performance Team the same day, which Jito quotes. SOLANA·HUB has not checked the measurement independently.
BAM is Jito’s Block Assembly Marketplace. The post defines a small block as a BAM block under 7 million compute units. The authors say that is not a definition for the whole Solana network.
The drop of about 77 percent is the move from about one block in 800 to one in 3,600. The authors name memory compaction in the Linux kernel, the kcompactd daemon, as the cause. The fix is two host settings: vm.compact_unevictable_allowed=0 and vm.compaction_proactiveness=0.
A second cut in the same post compares 16–29 September with 9–20 October. The share on BAM leaders fell from 0.1044% to 0.0325%. The post calls that 69 percent lower.
On leaders without BAM, the share in that window rose from 0.2911% to 0.5056%. Those figures are the team’s as well.
What to watch
- Whether a count outside Jito uses the same definition and the same windows