TDengine时序数据库性能测试
前言
TDengine Database官方及社区里有一些性能测试对比案例，不过发布的都比较早，其使用的版本都是早期低版本。本次测试参考官方提测的《使用 taosdemo 对 TDengine 进行性能测试》文章进行。因教程中的TDengine测版本较低，文章内容未进行更新，本次测试将使用2.6.0.28版本测试。
taosBenchmark 是TDengine专门为测试数据库开发的一个应用程序，用于对TDengine时序数据库进行写入和查询的性能测试，便于用户快速便捷的测试和体验TDengine高性能的特点。其工作逻辑是通过 taosBenchmark 模拟大量数据，还可以通过 taosBenchmark 参数灵活控制表的个数、表的列数、数据类型、一次插入的记录条数、记录的时间间隔、乱序比例、客户端接口类型以及并发线程数量等等。同时，taosBenchmark 也支持查询和订阅的测试。其运行操作也比较简单。本次我们针对其内置好的模拟数据进行写入和读取的性能测试。针对其更加丰富的参数使用，读者可自行去社区查阅学习。【本文仅供参考，有问题请留言讨论】
本次时序数据库的测试目标：
- TDengine 版本：2.6.0.28
前提条件
本次测试，使用以下环境
- 一台Linux服务器(因目前 2.X 版服务端 taosd 和 taosAdapter 仅在 Linux 系统上安装和运行)，包含10GB的空闲硬盘空间，用于存储产生的测试数据。本次测试采用的是华为云服务器：通用计算型 | s3.2xlarge.2 | 8vCPUs | 16GiB；CentOS 7.2 64bit。
- root权限。
准备测试
安装TDengine
TDengine 完整的软件包包括服务端（taosd）、用于与第三方系统对接并提供 RESTful 接口的 taosAdapter、应用驱动（taosc）、命令行程序 (CLI，taos) 和一些工具软件。
TDengine的安装有很多方式，包括：Docker、apt-get、安装包、源码方式进行安装，本次我们通过TDengine开源版本提供的tar.gz 格式安装包进行安装，可参考官方说明：安装和卸载 | TDengine 文档 | 涛思数据。
- 从官网下载获得 tar.gz 安装包，例如 TDengine-server-2.6.0.28-Linux-x64.tar.gz； 2、进入到 TDengine-server-2.6.0.28-Linux-x64.tar.gz 安装包所在目录，先解压文件后，进入子目录，执行其中的 ./install.sh -e no 安装脚本，如下是安装执行过程：
$ tar xvzf TDengine-enterprise-server-2.6.0.28-Linux-x64.tar.gz
TDengine-enterprise-server-2.6.0.28/
TDengine-enterprise-server-2.6.0.28/driver/
TDengine-enterprise-server-2.6.0.28/driver/vercomp.txt
TDengine-enterprise-server-2.6.0.28/driver/libtaos.so.2.6.0.28
TDengine-enterprise-server-2.6.0.28/install.sh
TDengine-enterprise-server-2.6.0.28/examples/
...
$ ll
total 43816
drwxrwxr-x 3 ubuntu ubuntu 4096 Feb 22 09:31 ./
drwxr-xr-x 20 ubuntu ubuntu 4096 Feb 22 09:30 ../
drwxrwxr-x 4 ubuntu ubuntu 4096 Feb 22 09:30 TDengine-enterprise-server-2.6.0.28/
-rw-rw-r-- 1 ubuntu ubuntu 44852544 Feb 22 09:31 TDengine-enterprise-server-2.6.0.28-Linux-x64.tar.gz
$ cd TDengine-enterprise-server-2.6.0.28/
$ ll
total 40784
drwxrwxr-x 4 ubuntu ubuntu 4096 Feb 22 09:30 ./
drwxrwxr-x 3 ubuntu ubuntu 4096 Feb 22 09:31 ../
drwxrwxr-x 2 ubuntu ubuntu 4096 Feb 22 09:30 driver/
drwxrwxr-x 10 ubuntu ubuntu 4096 Feb 22 09:30 examples/
-rwxrwxr-x 1 ubuntu ubuntu 33294 Feb 22 09:30 install.sh*
-rw-rw-r-- 1 ubuntu ubuntu 41704288 Feb 22 09:30 taos.tar.gz
$ sudo ./install.sh
Start to update TDengine...
Created symlink /etc/systemd/system/multi-user.target.wants/taosd.service → /etc/systemd/system/taosd.service.
Nginx for TDengine is updated successfully!
To configure TDengine : edit /etc/taos/taos.cfg
To configure Taos Adapter (if has) : edit /etc/taos/taosadapter.toml
To start TDengine : sudo systemctl start taosd
To access TDengine : use taos -h ubuntu-1804 in shell OR from http://127.0.0.1:6060
TDengine is updated successfully!
Install taoskeeper as a standalone service
taoskeeper is installed, enable it by `systemctl enable taoskeeper`
注意：install.sh 安装脚本在执行过程中，会通过命令行交互界面询问一些配置信息。如果希望采取无交互安装方式，那么可以用 -e no 参数来执行 install.sh 脚本。运行 ./install.sh -h 指令可以查看所有参数的详细说明信息。
本次测试我们直接用 -e no 参数来执行 install.sh 脚本。
- 启动
TDengine 使用 Linux 系统的 systemd/systemctl/service 来管理系统的启动和、停止、重启操作。TDengine 的服务进程是 taosd，默认情况下 TDengine 在系统启动后将自动启动。DBA 可以通过 systemd/systemctl/service 手动操作停止、启动、重新启动服务。
以 systemctl 为例，命令如下：
- 启动服务进程：systemctl start taosd
- 停止服务进程：systemctl stop taosd
- 重启服务进程：systemctl restart taosd
- 查看服务状态：systemctl status taosd
开展测试
在整个测试过程中，建议打开服务器的控制台监控，查看系统的CPU和内存占用情况。
测试过程
- 写入测试
- 测试用例1：
在不使用任何参数的情况下执行 taosBenchmark 命令，该命令将模拟生成一个电力行业典型应用的电表数据采集场景数据，即建立一个名为 test 的数据库，并创建一个名为 meters 的超级表，超级表的结构为：
taos> describe test.meters;
Field | Type | Length | Note |
=================================================================================
ts | TIMESTAMP | 8 | |
current | FLOAT | 4 | |
voltage | INT | 4 | |
phase | FLOAT | 4 | |
groupid | INT | 4 | TAG |
location | BINARY | 64 | TAG |
Query OK, 6 row(s) in set (0.002972s)
并按照 TDengine 数据建模的最佳实践，以 meters 超级表为模板生成一万个子表，代表一万个独立上报数据的电表设备，显示如下：
[root@ecs-3fb0 cfg]# taosBenchmark
[11/09 16:42:55.631713] INFO: taos client version: 2.6.0.28
Press enter key to continue or Ctrl-C to stop
[11/09 16:43:33.358768] INFO: create database: <CREATE DATABASE IF NOT EXISTS test PRECISION 'ms';>
[11/09 16:43:33.860524] INFO: stable meters does not exist, will create one
[11/09 16:43:34.361379] INFO: create stable: <CREATE TABLE test.meters (ts TIMESTAMP,current float,voltage int,phase float) TAGS (groupid int,location binary(24))>
[11/09 16:43:34.869963] INFO: generate stable<meters> columns data with lenOfCols<80> * prepared_rand<10000>
[11/09 16:43:34.880122] INFO: generate stable<meters> tags data with lenOfTags<62> * childTblCount<10000>
[11/09 16:43:34.884222] INFO: start creating 10000 table(s) with 8 thread(s)
[11/09 16:43:34.884813] INFO: thread[0] start creating table from 0 to 1249
[11/09 16:43:34.885088] INFO: thread[1] start creating table from 1250 to 2499
[11/09 16:43:34.885339] INFO: thread[2] start creating table from 2500 to 3749
[11/09 16:43:34.885700] INFO: thread[3] start creating table from 3750 to 4999
[11/09 16:43:34.887636] INFO: thread[4] start creating table from 5000 to 6249
[11/09 16:43:34.887926] INFO: thread[5] start creating table from 6250 to 7499
[11/09 16:43:34.888186] INFO: thread[6] start creating table from 7500 to 8749
[11/09 16:43:34.888430] INFO: thread[7] start creating table from 8750 to 9999
[11/09 16:43:48.838616] SUCC: Spent 13.9540 seconds to create 10000 table(s) with 8 thread(s), already exist 0 table(s), actual 10000 table(s) pre created, 0 table(s) will be auto create
然后 taosBenchmark 为每个电表设备模拟生成一万条记录，此时默认的：
– 每个写入请求记录数为30000，写入行设为10000；
– 写入线程数为8；
[11/09 16:43:51.824455] INFO: record per request (30000) is larger than insert rows (10000) in progressive mode, which will be set to 10000
[11/09 16:43:51.834750] INFO: Estimate memory usage: 11.82MB
Press enter key to continue or Ctrl-C to stop
[11/09 16:43:55.818914] INFO: thread[0] start progressive inserting into table from 0 to 1249
[11/09 16:43:55.818933] INFO: thread[3] start progressive inserting into table from 3750 to 4999
[11/09 16:43:55.818928] INFO: thread[2] start progressive inserting into table from 2500 to 3749
[11/09 16:43:55.818923] INFO: thread[1] start progressive inserting into table from 1250 to 2499
[11/09 16:43:55.818991] INFO: thread[4] start progressive inserting into table from 5000 to 6249
[11/09 16:43:55.819010] INFO: thread[5] start progressive inserting into table from 6250 to 7499
[11/09 16:43:55.819028] INFO: thread[6] start progressive inserting into table from 7500 to 8749
[11/09 16:43:55.821339] INFO: thread[7] start progressive inserting into table from 8750 to 9999
[11/09 16:44:25.821878] INFO: thread[3] has currently inserted rows: 8910000
[11/09 16:44:25.827670] INFO: thread[0] has currently inserted rows: 9230000
[11/09 16:44:25.831597] INFO: thread[4] has currently inserted rows: 9070000
[11/09 16:44:25.837201] INFO: thread[6] has currently inserted rows: 9050000
[11/09 16:44:25.839675] INFO: thread[7] has currently inserted rows: 8790000
[11/09 16:44:25.845131] INFO: thread[2] has currently inserted rows: 9540000
[11/09 16:44:25.852908] INFO: thread[1] has currently inserted rows: 9130000
[11/09 16:44:25.856755] INFO: thread[5] has currently inserted rows: 9410000
经过执行结果输出可以看到：
执行100000000条数据写入用时41秒，每秒钟插入性能高达 240 万 4104 条记录。官方提供的测试结果在同样的用例执行下（服务器配置信息未知）用了 6 秒的时间插入了一亿条记录，每秒钟插入性能高达 1628 万 5462 条记录。由于服务器资源配置和服务器写入延时较大的问题，导致本次测试结果与官方数据有一定的差别，属正常现象。但测试结果总体数据表现已经非常优秀。
[11/09 16:44:35.034487] SUCC: thread[2] completed total inserted rows: 12500000, 344858.35 records/second
[11/09 16:44:36.020174] SUCC: thread[1] completed total inserted rows: 12500000, 335522.93 records/second
[11/09 16:44:36.056786] SUCC: thread[5] completed total inserted rows: 12500000, 335212.72 records/second
[11/09 16:44:36.364985] SUCC: thread[0] completed total inserted rows: 12500000, 332542.39 records/second
[11/09 16:44:36.495492] SUCC: thread[6] completed total inserted rows: 12500000, 331381.29 records/second
[11/09 16:44:36.509238] SUCC: thread[4] completed total inserted rows: 12500000, 331320.34 records/second
[11/09 16:44:37.250942] SUCC: thread[3] completed total inserted rows: 12500000, 324860.17 records/second
[11/09 16:44:37.414442] SUCC: thread[7] completed total inserted rows: 12500000, 323742.61 records/second
[11/09 16:44:37.869222] SUCC: Spent 41.595522 seconds to insert rows: 100000000 with 8 thread(s) into test 2404104.94 records/second
[11/09 16:44:37.869252] SUCC: insert delay, min: 13.31ms, avg: 30.09ms, p90: 38.90ms, p95: 49.66ms, p99: 180.11ms, max: 358.02ms
- 测试用例2：
在<测试用例1>的测试过程中，将 taosBenchmark 命令带上 -T 48，意思为开启48个写入线程，执行过程如下：
[root@ecs-3fb0 cfg]# taosBenchmark -T 48
[11/09 17:26:13.519402] INFO: taos client version: 2.6.0.28
Press enter key to continue or Ctrl-C to stop
[11/09 17:26:15.134105] INFO: create database: <CREATE DATABASE IF NOT EXISTS test PRECISION 'ms';>
[11/09 17:26:15.635880] INFO: stable meters does not exist, will create one
[11/09 17:26:16.136620] INFO: create stable: <CREATE TABLE test.meters (ts TIMESTAMP,current float,voltage int,phase float) TAGS (groupid int,location binary(24))>
[11/09 17:26:16.657756] INFO: generate stable<meters> columns data with lenOfCols<80> * prepared_rand<10000>
[11/09 17:26:16.671521] INFO: generate stable<meters> tags data with lenOfTags<62> * childTblCount<10000>
[11/09 17:26:16.678598] INFO: start creating 10000 table(s) with 8 thread(s)
[11/09 17:26:16.679281] INFO: thread[0] start creating table from 0 to 1249
[11/09 17:26:16.679602] INFO: thread[1] start creating table from 1250 to 2499
[11/09 17:26:16.679915] INFO: thread[2] start creating table from 2500 to 3749
[11/09 17:26:16.682153] INFO: thread[3] start creating table from 3750 to 4999
[11/09 17:26:16.682462] INFO: thread[4] start creating table from 5000 to 6249
[11/09 17:26:16.682751] INFO: thread[5] start creating table from 6250 to 7499
[11/09 17:26:16.683011] INFO: thread[6] start creating table from 7500 to 8749
[11/09 17:26:16.683203] INFO: thread[7] start creating table from 8750 to 9999
[11/09 17:26:30.625863] SUCC: Spent 13.9470 seconds to create 10000 table(s) with 8 thread(s), already exist 0 table(s), actual 10000 table(s) pre created, 0 table(s) will be auto created
Press enter key to continue or Ctrl-C to stop
[11/09 17:26:31.526109] INFO: record per request (30000) is larger than insert rows (10000) in progressive mode, which will be set to 10000
[11/09 17:26:31.562217] INFO: Estimate memory usage: 51.99MB
Press enter key to continue or Ctrl-C to stop
[11/09 17:26:32.305446] INFO: thread[0] start progressive inserting into table from 0 to 208
[11/09 17:26:32.305473] [11/09 17:26:32.305459] INFO: thread[2] start progressive inserting into table from 418 to 626
[11/09 17:26:32.305449] INFO: thread[1] start progressive inserting into table from 209 to 417
INFO: thread[3] start progressive inserting into table from 627 to 835
[11/09 17:26:32.305515] INFO: thread[4] start progressive inserting into table from 836 to 1044
[11/09 17:26:32.305670] INFO: thread[15] start progressive inserting into table from 3135 to 3343
[11/09 17:26:32.306086] INFO: thread[18] start progressive inserting into table from 3760 to 3967
[11/09 17:26:32.307668] INFO: thread[44] start progressive inserting into table from 9168 to 9375
[11/09 17:26:32.307831] INFO: thread[30] start progressive inserting into table from 6256 to 6463
[11/09 17:26:32.307839] INFO: thread[45] start progressive inserting into table from 9376 to 9583
[11/09 17:26:32.307980] INFO: thread[17] start progressive inserting into table from 3552 to 3759
[11/09 17:26:32.307984] INFO: thread[36] start progressive inserting into table from 7504 to 7711
[11/09 17:26:32.308019] INFO: thread[8] start progressive inserting into table from 1672 to 1880
[11/09 17:26:32.308037] INFO: thread[40] start progressive inserting into table from 8336 to 8543
[11/09 17:26:32.308447] INFO: thread[41] start progressive inserting into table from 8544 to 8751
[11/09 17:26:32.308530] INFO: thread[6] start progressive inserting into table from 1254 to 1462
[11/09 17:26:32.308687] INFO: thread[19] start progressive inserting into table from 3968 to 4175
[11/09 17:26:32.308821] INFO: thread[38] start progressive inserting into table from 7920 to 8127
[11/09 17:26:32.309141] INFO: thread[37] start progressive inserting into table from 7712 to 7919
[11/09 17:26:32.308031] INFO: thread[27] start progressive inserting into table from 5632 to 5839
[11/09 17:26:32.310077] INFO: thread[21] start progressive inserting into table from 4384 to 4591
[11/09 17:26:32.310152] INFO: thread[20] start progressive inserting into table from 4176 to 4383
[11/09 17:26:32.310203] INFO: thread[31] start progressive inserting into table from 6464 to 6671
[11/09 17:26:32.308035] INFO: thread[39] start progressive inserting into table from 8128 to 8335
[11/09 17:26:32.310744] INFO: thread[46] start progressive inserting into table from 9584 to 9791
[11/09 17:26:32.310830] INFO: thread[11] start progressive inserting into table from 2299 to 2507
[11/09 17:26:32.311017] INFO: thread[29] start progressive inserting into table from 6048 to 6255
[11/09 17:26:32.311245] INFO: thread[47] start progressive inserting into table from 9792 to 9999
[11/09 17:26:32.311297] INFO: thread[24] start progressive inserting into table from 5008 to 5215
[11/09 17:26:32.311682] INFO: thread[12] start progressive inserting into table from 2508 to 2716
[11/09 17:26:32.312475] INFO: thread[23] start progressive inserting into table from 4800 to 5007
[11/09 17:26:32.312570] INFO: thread[32] start progressive inserting into table from 6672 to 6879
[11/09 17:26:32.312625] INFO: thread[22] start progressive inserting into table from 4592 to 4799
[11/09 17:26:32.312678] INFO: thread[42] start progressive inserting into table from 8752 to 8959
[11/09 17:26:32.308021] INFO: thread[25] start progressive inserting into table from 5216 to 5423
[11/09 17:26:32.313582] INFO: thread[35] start progressive inserting into table from 7296 to 7503
[11/09 17:26:32.313718] INFO: thread[43] start progressive inserting into table from 8960 to 9167
[11/09 17:26:32.308025] INFO: thread[26] start progressive inserting into table from 5424 to 5631
[11/09 17:26:32.314693] INFO: thread[28] start progressive inserting into table from 5840 to 6047
[11/09 17:26:32.314796] INFO: thread[9] start progressive inserting into table from 1881 to 2089
[11/09 17:26:32.314911] INFO: thread[33] start progressive inserting into table from 6880 to 7087
[11/09 17:26:32.315325] INFO: thread[16] start progressive inserting into table from 3344 to 3551
[11/09 17:26:32.315622] INFO: thread[14] start progressive inserting into table from 2926 to 3134
[11/09 17:26:32.315897] INFO: thread[34] start progressive inserting into table from 7088 to 7295
[11/09 17:26:32.315973] INFO: thread[5] start progressive inserting into table from 1045 to 1253
[11/09 17:26:32.316878] INFO: thread[10] start progressive inserting into table from 2090 to 2298
[11/09 17:26:32.317259] INFO: thread[7] start progressive inserting into table from 1463 to 1671
[11/09 17:26:32.317760] INFO: thread[13] start progressive inserting into table from 2717 to 2925
[11/09 17:27:00.365086] SUCC: thread[2] completed total inserted rows: 2090000, 75863.75 records/second
[11/09 17:27:00.493822] SUCC: thread[8] completed total inserted rows: 2090000, 75637.93 records/second
[11/09 17:27:00.825880] SUCC: thread[44] completed total inserted rows: 2080000, 74298.97 records/second
[11/09 17:27:01.045686] SUCC: thread[21] completed total inserted rows: 2080000, 73688.22 records/second
[11/09 17:27:01.049718] SUCC: thread[39] completed total inserted rows: 2080000, 73725.26 records/second
[11/09 17:27:01.054190] SUCC: thread[19] completed total inserted rows: 2080000, 73760.15 records/second
[11/09 17:27:01.091836] SUCC: thread[45] completed total inserted rows: 2080000, 73701.16 records/second
[11/09 17:27:01.110052] SUCC: thread[33] completed total inserted rows: 2080000, 73502.23 records/second
[11/09 17:27:01.157685] SUCC: thread[28] completed total inserted rows: 2080000, 73383.12 records/second
[11/09 17:27:01.194270] SUCC: thread[20] completed total inserted rows: 2080000, 73370.50 records/second
[11/09 17:27:01.227142] SUCC: thread[37] completed total inserted rows: 2080000, 73293.44 records/second
[11/09 17:27:01.261687] SUCC: thread[40] completed total inserted rows: 2080000, 73210.92 records/second
[11/09 17:27:01.353070] SUCC: thread[31] completed total inserted rows: 2080000, 72981.87 records/second
[11/09 17:27:01.366231] SUCC: thread[3] completed total inserted rows: 2090000, 73233.17 records/second
[11/09 17:27:01.371686] SUCC: thread[38] completed total inserted rows: 2080000, 72884.24 records/second
[11/09 17:27:01.426157] SUCC: thread[25] completed total inserted rows: 2080000, 72680.42 records/second
[11/09 17:27:01.456252] SUCC: thread[4] completed total inserted rows: 2090000, 73060.84 records/second
[11/09 17:27:01.558129] SUCC: thread[22] completed total inserted rows: 2080000, 72428.90 records/second
[11/09 17:27:01.559171] SUCC: thread[27] completed total inserted rows: 2080000, 72486.59 records/second
[11/09 17:27:01.569404] SUCC: thread[16] completed total inserted rows: 2080000, 72391.96 records/second
[11/09 17:27:01.589362] SUCC: thread[34] completed total inserted rows: 2080000, 72299.10 records/second
[11/09 17:27:01.589707] SUCC: thread[32] completed total inserted rows: 2080000, 72407.69 records/second
[11/09 17:27:01.594174] SUCC: thread[10] completed total inserted rows: 2090000, 72726.52 records/second
[11/09 17:27:01.610639] SUCC: thread[13] completed total inserted rows: 2090000, 72697.21 records/second
[11/09 17:27:01.631136] SUCC: thread[14] completed total inserted rows: 2090000, 72568.95 records/second
[11/09 17:27:01.636533] SUCC: thread[43] completed total inserted rows: 2080000, 72185.51 records/second
[11/09 17:27:01.648486] SUCC: thread[26] completed total inserted rows: 2080000, 72275.52 records/second
[11/09 17:27:01.675671] SUCC: thread[15] completed total inserted rows: 2090000, 72493.13 records/second
[11/09 17:27:01.775693] SUCC: thread[46] completed total inserted rows: 2080000, 71895.65 records/second
[11/09 17:27:01.802677] SUCC: thread[1] completed total inserted rows: 2090000, 72105.79 records/second
[11/09 17:27:01.829260] SUCC: thread[9] completed total inserted rows: 2090000, 72151.09 records/second
[11/09 17:27:01.880842] SUCC: thread[41] completed total inserted rows: 2080000, 71700.76 records/second
[11/09 17:27:01.911553] SUCC: thread[23] completed total inserted rows: 2080000, 71471.06 records/second
[11/09 17:27:01.982751] SUCC: thread[24] completed total inserted rows: 2080000, 71373.91 records/second
[11/09 17:27:02.063169] SUCC: thread[30] completed total inserted rows: 2080000, 71088.89 records/second
[11/09 17:27:02.063704] SUCC: thread[47] completed total inserted rows: 2080000, 71200.03 records/second
[11/09 17:27:02.064125] SUCC: thread[7] completed total inserted rows: 2090000, 71456.82 records/second
[11/09 17:27:02.094536] SUCC: thread[35] completed total inserted rows: 2080000, 71069.86 records/second
[11/09 17:27:02.151676] SUCC: thread[42] completed total inserted rows: 2080000, 70907.14 records/second
[11/09 17:27:02.152561] SUCC: thread[6] completed total inserted rows: 2090000, 71280.12 records/second
[11/09 17:27:02.163409] SUCC: thread[36] completed total inserted rows: 2080000, 71028.18 records/second
[11/09 17:27:02.163477] SUCC: thread[17] completed total inserted rows: 2080000, 70919.58 records/second
[11/09 17:27:02.174467] SUCC: thread[12] completed total inserted rows: 2090000, 71227.96 records/second
[11/09 17:27:02.197242] SUCC: thread[29] completed total inserted rows: 2080000, 70830.95 records/second
[11/09 17:27:02.206336] SUCC: thread[18] completed total inserted rows: 2080000, 70882.92 records/second
[11/09 17:27:02.235195] SUCC: thread[11] completed total inserted rows: 2090000, 71177.81 records/second
[11/09 17:27:02.258538] SUCC: thread[5] completed total inserted rows: 2090000, 71041.25 records/second
[11/09 17:27:02.294383] SUCC: thread[0] completed total inserted rows: 2090000, 70960.44 records/second
[11/09 17:27:02.751293] SUCC: Spent 29.988757 seconds to insert rows: 100000000 with 48 thread(s) into test 3334583.02 records/second
[11/09 17:27:02.751336] SUCC: insert delay, min: 14.20ms, avg: 138.17ms, p90: 194.66ms, p95: 217.56ms, p99: 265.10ms, max: 374.39ms
经过执行结果输出可以看到：
执行100000000条数据写入用时29.9秒，每秒钟插入性能高达 333 万 4583 条记录。
- 查询测试
- 测试用例3：
taosBenchmark 给出了查询测试的方法，是提供了一个 -x 参数，可以在插入数据结束后进行常用查询操作并输出查询消耗时间。测试时是通过<测试用例1>预先写入一亿条数据，然后对这一亿条数据进行查询性能测试，查询的测试主要对常用聚合函数进行操作，包括：count()、avg()、sum()、max()、min()、first()、last()等。测试过程如下图：
通过测试结果可以看到：通过查询语句取出一亿条记录操作仅消耗2.1秒。而对一亿条记录进行常用的聚合函数操作通常仅需要几十毫秒，时长最快的仅十几毫秒，时间最长的 count 函数也才到一百多毫秒。查询性能表现非常优秀。
结果分析
通过本测试产生的数据和相关的写入、查询用例测试可以看出，TDengine在性能上表现非常优秀。
细致分析下来可以有以下结论【在8核CPU的服务器下】：
- 写入性能：执行100000000条数据写入用时29.9秒，每秒钟插入性能高达 333 万 4583 条记录。
- 普通查询全部数据：通过查询语句取出一亿条记录操作仅消耗2.1秒。
- 全部数据聚合计算查询：对一亿条记录进行常用的聚合函数操作通常仅需要几十毫秒，时长最快的
仅十几毫秒，时间最长的 count 函数也才到一百多毫秒。
通过华为云控制台监控的观察，测试用例执行时，TDengine的CPU占用率仅达到44.6%【没有更详细的测试插件，仅供参考】，表明在8核CPU的服务器下TDengine的CPU占用率低很多。
在实际测试过程中，分别测试过8线程、12线程、16线程、48线程…192线程等。结果发现当线程数等于服务器核数的2倍时，其测试结果最佳。当线程数大于CPU核数2倍以上时，测试结果表现出其写入用时基本处于平稳。