mirror of
https://github.com/scylladb/scylladb.git
synced 2026-04-22 01:20:39 +00:00
"This series optimizes rows for cases when the number of cells is small. Instead of storing cells always in an intrusive set which isn't cache friendly and have big memory overhead either managed_vector is used are used, provided that the cell count is below cetrain threshold. Fixes #130. Below are the results of perf_mutation and perf_simple_query, medians of 20 results are used. perf_mutation (-c 1): before: 3216314.09 tps after: 3375013.94 tps diff: +4.93% perf_simple_query (read): before: 535327.27 tps after: 541866.17 tps diff +1.22% perf_simple_query (write): before: 707144.21 tps after: 749960.52 tps diff: +6.05%"