~/talk

MatrixKV: Reducing Write Stalls and Write Amplification in LSM-tree Based KV Stores

Talkat USENIX ATC '20 · 2020-07-15..07-17· [databases]
name:Ting Yao
links:none on file
links:none on file
links:none on file
name:Qiu Cui
links:none on file
name:Liu Tang
links:none on file
links:none on file
links:none on file
name:Xubin He
links:none on file
slides:none on file

MatrixKV implements a matrix container in NVM (Optane DC PMM) inside RocksDB, adds column compaction and cross-row hint search, and flattens levels to cut write amplification. It reduces each L0→L1 compaction to about 33 MB compaction (vs ~3.1 GB in RocksDB and 15 GB in NoveLSM), lowers write amplification by ~3.4×, and delivers 3.6×/2.6× higher random write throughput and 5×/1.9× lower 99th‑percentile latency compared to RocksDB and NoveLSM.

AI-written summary · about

  1. 0:00Intro
  2. 0:31Outline
  3. 1:56Challenge 1: Write stall
  4. 3:29Root cause of write stall: LO-L1 compaction
  5. 5:27Root cause of increased write amplification
  6. 7:10Motivation
  7. 8:02Overall Architecture
  8. 9:12Matrix Container
  9. 10:21Row Table
  10. 11:10Fine grained column compaction
  11. 13:19Reducing LSM-tree depth
  12. 14:08Cross-Row hint search
  13. 16:01Evaluation Setup
  14. 16:37Random Write Throughput
  15. 17:25Write stalls
  16. 17:59Tail Latency
  17. 18:25Fine granularity column compaction
~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.