Benchmarks 13-Way VPS Server Benchmark Comparison Tests Discussion Thread

This thread is dedicated to discussing my 13-way VPS benchmark comparison test results.

---

Got questions or even share your own experiences with those specific VPS provider's plans

 

FYI, as noted in the vps benchmark comparison review thread, anyone can replicate the exact benchmark testing I did with a fresh virgin CentOS 7.6+ 64bit server (whether it's VPS or baremetal dedicated server based) using my installnbench2.sh script. Depending on speed of your server, such installnbench2.sh runs can take between 1.5-6+ hrs to run.

Running installnbench2.sh

All VPS servers will run CentOS 7.6 64bit latest after a yum update and server reboot to ensure all VPS servers are on the same page in terms of updates and Linux Kernel updates for Meltdown/Spectre variant and MDS/Zombieload vulnerability patch fixes which affect performance.

Code (Text):

yum -y update && reboot

Then running installnbench2.sh script in a screen session so SSH disconnections do not abort the benchmark tests.

Code (Text):

mkdir -p /root/tools
cd /root/tools
yum -y install wget
wget -O installnbench2.sh https://github.com/centminmod/centminmodbench/raw/master/installnbench2.sh
chmod +x installnbench2.sh
yum -y install screen
screen -dmS installnbench
screen -r installnbench
time /root/tools/installnbench2.sh

If you do get SSH session disconnected during testing, just log back in and reattach the screen session to see the benchmark that had continued to run even when SSH session was disconnected via command

Code (Text):

screen -r installnbench

On completion run of benchmarks from installnbench2.sh, you can collect all the Centmin Mod install and benchmark logs via tar command below which will create a tar'd gzip compressed file at /svr-setup/all-benchmark-logs.tar.gz .

Code (Text):

cd /svr-setup
tar -I pigz -cvf /svr-setup/all-benchmark-logs.tar.gz /home/centminmodbench_logs /root/centminlogs /home/sysbench

Logs are mainly where the benchmark results are and 1 log is the Centmin Mod LEMP stack install log. These logs have date/timestamps appended to the end so these are just examples

• For installnbench2.sh all centminmodbench.sh benchmark raw results which include the zcat/pzcat and geekbench results at /home/centminmodbench_logs/centminmodbench_results_130619-084257.log and /home/centminmodbench_logs/_publicreport_130619-084257.log - compacted results format with preformatted forum CODE bbcode tags wrapped around the results for easier posting/sharing on forums.
• For Centmin Mod Nginx HTTP/2 HTTPS benchmarks - /root/centminlogs/https-benchmark-all-130619-095805.log which show raw results including how the dummy test Centmin Mod Nginx HTTP/2 HTTPS dual RSA/ECDSA SSL certificated supported vhost site was created and /root/centminlogs/h2load-nginx-https-130619-095805.log which just shows the results only for nghttp2's h2load HTTP/2 HTTPS load tester.
• For Redis benchmark results /root/centminlogs/centminmod-benchmark-redis-tests-130619-101110.log and also markdown formatted table results within /root/centminlogs/centminmod-benchmark-redis-tests-130619-101110-markdown-*.log logs
• For Centmin Mod Nginx HTTP/2 HTTPS vhost creation log via nv SSH commandline method of generating Nginx vhosts, /root/centminlogs/centminmod_130619-095806_nginx_addvhost_nv.log
• For all the sysbench.sh benchmark results saved in directory at /home/sysbench. There's 100s of logs created of which the ones in tabulated markdown format are labeled as sysbench-*-markdown.log after the sysbench test done. For example for sysbench 2 cpu thread test labeled as sysbench-cpu-threads-2-markdown.log while non-markdown formatted results are in log file without markdown suffix, sysbench-cpu-threads-2.log
• For Centmin Mod LEMP stack install log, /root/centminlogs/centminmod_123.09beta01.b185_130619-102043_install.log with raw results, /root/centminlogs/installer_130619-101111.log with raw results + breakdown of install time metrics and /root/centminlogs/installer_130619-101111_minimal.log which is compact results with some of the result verbose output stripped to result size of log and number of lines in log.

 

much appreciated mate

Less of a concern especially with 500 concurrent users if you have Cloudflare CDN in front of your Centmin Mod Nginx. Though depends on overall throughput traffic your site does.

 

@BigIron just asked Upcloud for clarification and all their VPS servers use 1Gbit/s public network links now not 500Mbit/s

 

much appreciated @Kenny

 

Thanks for the kind words and suggestion. Internal networking has been on my to do list though there's so many factors and differences in how it's setup with various web hosts to make it comparable. Like for me, I always configure more specific CSF Firewall rules for internal LAN IPs which also factor into the equations. Probably would be easier to actually test MySQL workloads when a remote MySQL server is in play instead of the network layer level.