LiteSpeed vs. Nginx vs. Apache
In this benchmark, we compare the performance of LiteSpeed Web Server to Nginx and Apache both with and without cache. We measured averages for these servers across three statistics: requests handled per second, maximum transaction time in seconds, and throughput in megabytes per second. These three factors are all important parts of server performance that influence stability, performance, and scalability.
SSL Requests Per Second with Parser
Number of PagesThis graph shows the effects that SSL has on web pages, with caching enabled and each page being parsed, with each user visiting 1-10 & 100 pages before leaving. After visiting 10 pages, we see that LiteSpeed Enterprise + LiteMage Cache performs 14% faster than LiteSpeed Enterprise 5.1.9 + LiteMage 1.2.4, 62% faster than Apache + Varnish, and 102% faster than Nginx + Varnish.
SSL Requests Per Second without Parser
Number of PagesThis graph shows the effects that SSL has on web pages, with caching enabled and each page being parsed, with each user visiting 1-10 & 100 pages before leaving. After visiting 10 pages, we see that LiteSpeed Enterprise + LiteMage Cache performs 10% faster than LiteSpeed Enterprise 5.1.9 + LiteMage 1.2.4, 266% faster than Apache + Varnish, and 514% faster than Nginx + Varnish.
Load Average Per Page
Number of PagesThis graph shows the effects that SSL has on average server load when requesting web pages, with caching enabled and each page being parsed, with each user visiting 1-10 & 100 pages before leaving. After visiting 10 pages, we see that LiteSpeed Enterprise + LiteMage Cache has 45% lower load than LiteSpeed Enterprise 5.1.9 + LiteMage 1.2.4, 240% lower load than Apache + Varnish, and 172% lower load than Nginx + Varnish.
Load Average
This graph shows the effects that SSL has on average server load when requesting web pages, with caching enabled and each page being parsed, with each user visiting 1-10 & 100 pages before leaving. Over the course of 10 minutes, LiteSpeed Enterprise + LiteMage Cache has 42% lower load than LiteSpeed Enterprise 5.1.9 + LiteMage 1.2.4, 139% lower load than Apache + Varnish, and 227% lower load than Nginx + Varnish.
Requests Per Second
Number of PagesThis graph shows the number of requests per second handled when requesting web pages, with caching enabled, with each user visits 1-10 & 100 pages before leaving. After visiting 10 pages, we see that LiteSpeed Enterprise + LiteMage Cache performs 153% faster than LiteSpeed Enterprise 5.1.9 + LiteMage 1.2.4, 3228% faster than Apache + Varnish, and 4892% faster than Nginx + Varnish.
Turpentine Cookie Bypass
Turpentine has a built-in "benchmark" mode that treats any requests detected as coming from a benchmarking tool as a crawler request, hard coding the frontend cookie to 'frontend=crawler-session'. This mode will use the same cached private blocks for all page requests, regardless of who is actually visiting the page, completely avoiding the backend by breaking Magento functionality. This was likely implemented to get better benchmarking performance than is actually possible by Varnish in a real word scenario.
LiteMage Cache on the other hand is optimized to completely avoid the PHP backend for catalog browsing without breaking any Magento functionality. Even with Varnish's unrealistic "benchmark" mode, LiteSpeed Enterprise + LiteMage Cache is still 67% faster than Varnish + Apache, and 180% faster than Varnish + Nginx.
Notes:
- Opcode caching was used for all setups.
- The test was performed over at Digital Ocean all in NYC1 Datacenter.
Test Environment
Software Configuration:
Apache 2.4.23 Nginx 1.11.5 LSWS 5.1.10 Varnish 4.1.3 + Turpentine 0.7.0 LiteMage 1.3.1 LSPHP 7.0.12 PHP-FPM 7.0.12 MariaDB 10.1.18
Server hardware specs:
1 Cores from an Intel(R) Xeon(R) CPU E5-2650L v3 @ 1.80GHz 1GB RAM 30GB SSD CentOS 7.2
Client hardware specs:
2 Cores from an Intel(R) Xeon(R) CPU E5-2650L v3 @ 1.80GHz 2GB RAM 40GB SSD CentOS 7.2
We welcome your feedback on our forum.