DM 1.0-alpha Benchmark Report

This DM benchmark report describes the test purpose, environment, scenario, and result.

Test purpose

The purpose of this test is to test the performance of DM incremental replication.

Note

The results of the testing might vary based on different environmental dependencies.

Test environment

Machine information

System information:

Machine IP	Operation system	Kernel version	File system type
192.168.0.6	CentOS Linux release 7.6.1810	3.10.0-957.1.3.el7.x86_64	ext4
192.168.0.7	CentOS Linux release 7.6.1810	3.10.0-957.1.3.el7.x86_64	ext4
192.168.0.8	CentOS Linux release 7.6.1810	3.10.0-957.1.3.el7.x86_64	ext4
192.168.0.9	CentOS Linux release 7.6.1810	3.10.0-957.1.3.el7.x86_64	ext4
192.168.0.10	CentOS Linux release 7.6.1810	3.10.0-957.1.3.el7.x86_64	ext4
192.168.0.11	CentOS Linux release 7.6.1810	3.10.0-957.1.3.el7.x86_64	ext4

Hardware information:

Type	192.168.0.9, 192.168.0.10, 192.168.0.11	192.168.0.6, 192.168.0.7, 192.168.0.8
CPU	8 vCPUs, Intel(R) Xeon(R) Platinum 8163 CPU @ 2.50GHz	4 vCPUs, Intel(R) Xeon(R) Platinum 8163 CPU @ 2.50GHz
Memory	16G	8G
Disk	1T Aliyun ESSD	256G Aliyun ESSD
Network card	1 Gigabit Ethernet, 1000Mb/s	1 Gigabit Ethernet, 1000Mb/s

Cluster topology

Machine IP	Deployment instance
192.168.0.9	TiKV * 1, TiDB * 1
192.168.0.10	TiKV * 1
192.168.0.11	TiKV * 1
192.168.0.6	PD * 1, MySQL * 1, DM-worker * 1
192.168.0.8	PD * 1, MySQL * 1, DM-worker * 1
192.168.0.7	PD * 1, DM-master * 1

Version information

MySQL version: 5.7.25-log
TiDB version: v3.0.0-beta-27-g6398788
DM version: v1.0.0-alpha-10-g4d01d79
Sysbench version: 1.0.9

Test scenario

Data flow

MySQL1 (192.168.0.8) -> DM-worker1 (192.168.0.6) -> TiDB (192.168.0.9)

Test procedure

Set up environment
Use sysbench to create the table and generate the initial data in upstream MySQL
Start DM-task in the all mode
Use sysbench to generate incremental data in upstream MySQL

Use sysbench to generate data load in upstream MySQL

Upstream sysbench test script:

sysbench --test=oltp_insert --tables=2 --num-threads=1024 --mysql-host=192.168.0.8 --mysql-port=3306 --mysql-user=root --mysql-db=dm_poc --db-driver=mysql --report-interval=10 --time=900 run

The structure of the table used for the test:

CREATE TABLE `sbtest` (
  `id` int(11) NOT NULL AUTO_INCREMENT,
  `k` int(11) NOT NULL DEFAULT '0',
  `c` char(120) CHARSET utf8mb4 COLLATE utf8mb4_bin NOT NULL DEFAULT '',
  `pad` char(60) CHARSET utf8mb4 COLLATE utf8mb4_bin NOT NULL DEFAULT '',
  PRIMARY KEY (`id`),
  KEY `k_1` (`k`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin

The deployment and configuration details

// TiKV configuration
sync-log = false
[defaultcf]
block-cache-size = "4GB"
[writecf]
block-cache-size = "4GB"
[raftdb.defaultcf]
block-cache-size = "4GB"

// The task configuration of the sync processing unit in DM
syncers:
    worker-count: 256
    batch: 100
    max-retry: 20

Test result

items	threads	qps	tps	95% Latency (ms)
MySQL	1024	15.10k	15.10k	121.08
DM	256	13.89k	13.89k	210 (txn execution latency)
TiDB	-	18.53k (Begin/Commit 2.4k Replace 13.80k)	2.27k	29