[CASSANDRA-21083][trunk] Optimize memtable flush logic#4536

Open

netudima wants to merge 12 commits intoapache:trunkfrom

netudima:CASSANDRA-21083-trunk

Commits on Dec 20, 2025

add more flushing stats: partitions/rows, bytes rate, CPU and heap allocation for the flushing thread
netudima
committed
invoke metadataCollector.updateClusteringValues only for first and last clustering key in a partition
netudima
committed
split call sites for in Cell serialize logic, make isCounterCell cheaper (avoid megamorphic call + cache isCounterColumn)
netudima
committed
check if Guardrails enabled at the beginning of writing, avoid hidden auto-boxing for logging of primitive parameters
netudima
committed
add fast return for BTreeRow.hasComplexDeletion, avoid column.name.bytes.hashCode if not needed, avoid capturing lambda allocation in UnfilteredSerializer.serializeRowBody
netudima
committed
reduce allocations during serialization of NativeClustering
netudima
committed
do not re-map colums in serializeRowBody if they haven't changed
netudima
committed
add flushing iterator without columns filtering
netudima
committed
improve cell methods inlining and try to use monomorphic calls
netudima
committed
minor changes in completed flushing log message
netudima
committed
add columnsChangedAfterCreation implementation for SkipListMemtable and ShardedSkipListMemtable
netudima
committed

Commits on Dec 21, 2025

address review comments
netudima
committed