Shardingsphere: [Performance Test] Provide a full-scale and multi-scenario benchmark test

Created on 30 Jul 2020  ·  1Comment  ·  Source: apache/shardingsphere

Hi Community,

Background

I know performance test is your concern when considering a new solution for your systems or projects. Recently we received many feedbacks and complaints about our benchmark graphs.

After a rough discussion, we decide to commit more energy to the performance test of Apache ShardingSpehre and provide a full-scale and multi-scenario benchmark test for the subsequent releases.

Content

  1. The comparison among various releases.
  2. The comparison among ShardingJDBC, ShardingProxy, and MySQL.
  3. The comparison among Sharding feature, Encrypt feature, MasterSlave feature.
  4. Different query SQLs for the above.
  5. Benchmark UI

Project

shardingsphere-benchmark

test performance refactor

Most helpful comment

Recently i have been investigating ShardingProxy and ShardingJDBC performance.

Senarios i did

  • Base Data 1000w vs Base Data 1000, observe tps of ShardingDBC & ShardingProxy;
  • With base data 1000w, comparing ShardingJDBC&ShardingProxy to Mysql, observe tps of ShardingDBC & ShardingProxy and Mysql;

Result i got

  • At senario 1 - base data 1000w vs base data 1000, ShardingDBC & ShardingProxy tps improves a lot with base data 1000w, refer to image 1 as detail data ;
  • At senario 2 - with base data 1000w, comparing ShardingJDBC&ShardingProxy to Mysql, their tps is much better for full routing select/full routing insert+update+delete/masterslave+encrpytinsert+update+delete than mysql. In addition their tps is lower than Mysql, refer to image 2 as detail data ;

Senario 1
image

Senario 2
image

>All comments

Recently i have been investigating ShardingProxy and ShardingJDBC performance.

Senarios i did

  • Base Data 1000w vs Base Data 1000, observe tps of ShardingDBC & ShardingProxy;
  • With base data 1000w, comparing ShardingJDBC&ShardingProxy to Mysql, observe tps of ShardingDBC & ShardingProxy and Mysql;

Result i got

  • At senario 1 - base data 1000w vs base data 1000, ShardingDBC & ShardingProxy tps improves a lot with base data 1000w, refer to image 1 as detail data ;
  • At senario 2 - with base data 1000w, comparing ShardingJDBC&ShardingProxy to Mysql, their tps is much better for full routing select/full routing insert+update+delete/masterslave+encrpytinsert+update+delete than mysql. In addition their tps is lower than Mysql, refer to image 2 as detail data ;

Senario 1
image

Senario 2
image

Was this page helpful?
0 / 5 - 0 ratings