Skip to content

Latest commit

 

History

History
56 lines (43 loc) · 2.6 KB

2016-07-23-tidb-weekly.md

File metadata and controls

56 lines (43 loc) · 2.6 KB
Raw
title date summary tags
Weekly update (July 17 ~ July 22, 2016)
2016-07-23
Last week, we landed 22 PRs in the TiDB repositories and 15 PRs in the TiKV repositories.
TiDB
TiKV
Placement Driver

Weekly update (July 17 ~ July 22, 2016)

Last week, we landed 22 PRs in the TiDB repositories and 15 PRs in the TiKV repositories.

Notable changes to TiDB

  • Refactor the query optimizer to imporve the query efficiency for Join and SubQuery
  • Add distributed SQL support for aggregate functions
  • Improve the stability of the TiDB service
  • Refactor the Decimal codes to improve the compatibility with MySQL
  • Optimize the TiDB compatibility and performance for Zabbix
  • Enhance the performance and the Sysbench result is improved significantly

Notable changes to TiKV

  • Add asynchronous scheduler support for higher throughput and better performance, see Benchmark.
  • Use PipeBuf to speed up socket read/write.
  • Add pushing down max/min support for the coprocessor.
  • Re-use RocksDB write ahead log (WAL) to guarantee consistency when writing data in different column families.
  • Support using make install on the CentOS platform to install TiKV.

Notable changes to Placement Driver

  • Refactor the balance framework to make a cluster more balanced and stable.
  • Support web UI in Docker.

Benchmark

Use sysbench to benchmark asynchronous scheduler and previous 8 threadpools in 3-node TiKV.

# Prepare data
sysbench --test=./lua-tests/db/oltp.lua --mysql-host=${host} --mysql-port=${port} \
 --mysql-user=${user} --mysql-password=${password} --oltp-tables-count=$1 \
 --oltp-table-size=1000 --rand-init=on prepare

# Run benchmark
sysbench --test=./lua-tests/db/insert.lua --mysql-host=${host} --mysql-port=${port} \
 --mysql-user=${user} --mysql-password=${password} --oltp-tables-count=1 \
 --oltp-table-size=1000 --num-threads=${threads} --report-interval=60 \
 --max-requests=1280000 --percentile=99 run

|Threads|Async scheduler qps|Async scheduler avg/.99 latency|8 threadpools qps|Thread pool avg/.99 latency| |---|---|---|---|---|---| |32|2049|15.6/29.1|1652|19.4/36.3| |64|2042|31.3/85.5|1693|37.8/83| |128|2125|60.2/147|1649|77/175|

As we can see, the qps is increased by about 25%, and the latency is decreased by about 15%.