The web-server fight of the century continues through the pandemic into 2022.
- Who’s faster?
- Is speed all that matters?
- What are the experts choosing?
Find out as they go head-to-head in another of my lazy ass write-ups. Just a couple photos, biased opinions, and hasty conclusions from my 30-minute testing of these two server giants.
Who are NGINX and LiteSpeed?
The biggest and most popular high performance web servers in use today.
If you seriously haven’t heard of them, you’re probably living under a rock and should get an internet subscription.
Not even kidding, these 2 names are mentioned everywhere today. Back in the days…it was only Apache and NGINX…and not much of a battle (speedwise) because NGINX was way newer and built specifically for speed. But Apache hung on because it could do a dozen things NGINX couldn’t.
LiteSpeed (2003) actually released a year before NGINX (2004) and was fast like NGINX but could also do all the things Apache did. You’d think LiteSpeed would be mega-popular but it wasn’t because you had to pay for it…whereas Apache and NGINX were free and open-source. After some years, LiteSpeed decided to release a free open-source version called OpenLiteSpeed (2013).
Which takes us to today’s webhosting landscape:
- Many traditional hosts still offering Apache.
- Newer modern hosts offering NGINX.
- Boutique hosts (both traditional & modern) offering LiteSpeed.
- Ultra-modern trendy hosts offering NGINX & OpenLiteSpeed (OLS).
NGINX vs OpenLiteSpeed (OLS)
So now we realize where the battle is really at. It’s free open-source OpenLiteSpeed taking on the powerhouse free open-source NGINX.
Questions the noobs asks:
- Is OLS really faster than NGINX? – that’s what my post will answer.
- Is OLS newer/cooler than NGINX? – yes, it is.
- Is OLS easier to use than NGINX? – yes, it is. In terms of htaccess compatibility and graphical UI configuration console.
- Main advantage OLS has over NGINX? – OLS allows you to use LiteSpeed Cache plugin (which is my favorite cache plugin for WordPress).
Questions the pros ask:
- Is OLS really faster than NGINX? – diehard NGINX devs/admins say no. Diehard LS devs/admins say yes.
- Is OLS newer/cooler than NGINX? – some pros think it’s gimmicky and not as proven, not as stable.
- Is OLS easier to use than NGINX? – not if they prefer NGINX config.
- Advantages NGINX has over OLS? – admins today are more familiar with NGINX config, and find it less hassle to config/manage.
- Advantages OLS has over NGINX? – easy learning curve for pros, easy interface for non-pros, htaccess compatible, also the super powerful native WordPress cache plugin (LiteSpeed Cache) whereas NGINX doesn’t have one.
NGINX vs OLS speed benchmark charts
Some notes about my testing:
- I fired up a bunch of cheap $5 VPS (1-core, 1gb ram). One with NGINX, another with OLS. (Using mostly default configs, I didn’t fine-tune them.)
- NGINX server had fresh WP site with only Cache Enabler plugin.
- OLS server had fresh WP site with only LiteSpeed Cache plugin.
- Then used free loader.io account to throw 10,000 clients/min at HTTPS domain using GET method.
NGINX (un-cached) vs OLS (un-cached)
What this means:
- NGINX served 150 requests and averaged 1-sec for each, before failing in 6 seconds.
- OLS served 383 requests but averaged 9-sec for each, before failing in 16 seconds.
- Raw OLS is simpler weaker to me. I much prefer raw NGINX handling fewer requests at reasonable response (1sec) before crashing early, than OLS with slower initial response before trying to unsuccessfully handle more concurrent hits.
- But realistically, this comparison doesn’t matter. Because nobody in their right mind would be serving 10k requests/min without caching.
NGINX (cached) vs OLS (cached)
What this means:
- Don’t bother looking at average response times. Both servers run virtually at the same speed once the cache kicks in after 2 seconds (~66ms per request). The average is mostly affected by the slow initial response being averaged out.
- But indeed, OLS is faster by a tiny tiny margin. Initial slow response was quicker (OLS 106ms vs NGINX 137ms) and then the stable response was a tiny bit faster (OLS 65ms vs NGINX 67ms).
So is OpenLiteSpeed truly faster/better than NGINX?
Is OpenLiteSpeed faster than NGINX?
- Without caching, I’d say NO.
- But with caching, I’d say YES but only by a freaken hair. Difference so small you wouldn’t even notice.
Any other benchmarks showing LiteSpeed being 20-300% faster than NGINX are silly to me. I’m a diehard LiteSpeed fan myself but don’t believe it’s that much faster. OpenLiteSpeed (free version) is like 1-2% faster at most. LiteSpeed Enterprise (paid version) is maybe 3-5% faster at most. If even that. And sometimes, I swear NGINX is faster.
However…
LiteSpeed can do many things that produce faster end result in real-world use. Caching dynamic requests, security features (that decrease resource drain from DDOS attacks), and more.
What’s the real benefit of LiteSpeed/OpenLiteSpeed over NGINX?
The #1 reason for me is the free LiteSpeed Cache plugin (compatible with only LS/OLS servers). It’s a freaken amazing plugin with dozens of performance features to speed up your site in so many ways. This plugin alone is what separates LS/OLS from NGINX. NGINX doesn’t have any native cache plugin for WordPress. Sure, you can use other cache plugins on NGINX but they aren’t as powerful as the LiteSpeed Cache plugin.
Of course, NGINX diehards will argue that NGINX can do this too. They’ll claim “NGINX also has micro-caching capabilities to rival LiteSpeed’s ESI dynamic cache”. Which is yeah, it’s true, but it’s not as easy to configure for the everyday person. With LiteSpeed cache plugin, you just make a few clicks. With NGINX, you gotta learn how to make NGINX configs and manually write out exactly what you need. Yeah…no thanks. Just not realistic unless you’re an NGINX expert managing only a few sites.
Let’s not forget about LiteSpeed’s QUIC.cloud service (which easily integrates with LiteSpeed Cache plugin) and caches your site at the edge…rivaling cloud-cache services like NitroPack for a fraction of the price. NGINX doesn’t have anything like that.
How urgently should you switch to OpenLiteSpeed?
Well, the majority of high-performance web servers out there are still happily running on NGINX. It works well, it’s proven, can still be cached, and many admins are familiar with it. You’ll still get great performance out of it.
But if you’re curious to try something new…OpenLiteSpeed is worth looking at. For that reason, you may have noticed many new hosts adopting it.
- If you’re happily on NGINX, don’t sweat it.
- If you’re shopping for a new server, LiteSpeed/OpenLiteSpeed are worth a look.
- If your name is WPJohnny, you’ve only used LiteSpeed for the past 10+ years. 😉
My site is hosted on a Slow shared server but it has LiteSpeed server and the fully load time is around 300MS and I have been using LiteSpeed for over 4 years….it is NameCheap…
I may be 3 years late for replying but, cmon, its shared hosting and namecheap is known to cramp up clients on one server.
I used webinoly, centminmod, Easyengine and wordops, a fine tuned Nginx stack for WordPress but now using Litespeed enterprise shared hosting for my personal blog. Nginx is great but I have to buy some plugin to make WP extra faster or used 3 to 4 free plugin to achieve the same results. (Caching, Minification, Lazy loading, etc)
Litespeed caching plugin does everything like those paid plugin, even I’ll able to generate ucss for site using free quick.cloud account, which your unable to generate if you won’t buy any premium plugin in nginx case.
Yes, exactly. LiteSpeed Cache plugin has so many features that cost money in other plugins and also you’d probably need several to do everything it can. Now whether you NEED all those features is a different conversation altogether.
Truly agreed with you. Having it’s own dedicated LSCache plugin together with Litespeed server, it is the best combo!
How about Apache+NGINX vs OLS?
Still a tiny difference?
I would most definitely pick OLS over Apache+NGINX… since the main reason why people use that hybrid combo is to combine Apache/htaccess capabilities with NGINX performance. And OLS was already built with that in mind from the very beginning (performance + Apache compatibility).
Thanks
“NGINX vs OpenLiteSpeed (OLS) honest speed comparison 2022”
Seems like we have a bright future to look forwards to in 2022 .. =)
Great write-up, concludes my findings to, thanks Johnny
I’ve tried several different hosts running on either Open LiteSpeed and LiteSpeed Enterprise. Although my site felt faster than when I was with WPX (Apache) the hassle of properly setting up the page cache (using the LSCache plugin) for an e-commerce platform other than WooCommerce (Easy Digital Downloads in my case) is — in a word: insane!
Played with it for about 3 weeks, fixed one issue when another arose. Lost a lot of money in the process — aaaargh!
But, in all honesty, their support is top notch. They were very helpful. Now I’m with a speedy host running a custom, tweaked Linux build with Nginx and like you said: it’s much more stable than LS and I don’t “feel” the difference in speed compared to LS.
Ooooh….wish I had a chance to see your LSC setup. But glad you found a happy home eventually!
I experienced the same issues on ols inclusive i used a managed vps and problems and more problems on woocommerce.
What tweakes Linux nginx do you use? If you can share would be great i am thinking on try nginx on a vps too.
Thanks you good luck!!!
Hi Ivan,
Back when I wrote that comment I was with Servebolt. Great hosting provider, but things went wrong in my case a few times too much (they updated my Linux image, without prior notice, breaking things, adding services accidentally, etc.) So, I’d say they have a lot of growing to do professionally.
Last January I migrated to Kinsta, and I never looked back. Skilled support staff, a wide range of tools and speedy servers. Haven’t had any issues since I joined them.
Definitely agree with some points mentioned in this article
1. Nginx when properly configured is better than out of box Nginx defaults. FYI, Cache-Enabler WP plugin has 2 methods of doing full HTML page caching. The default which still relies on PHP for the logic. Or via advanced caching to totally offload the cache logic + serving to Nginx bypassing PHP. Excample of what Nginx + Cache Enabler WP plugin with advanced caching can do – benchmarks at bottom of blog post https://blog.centminmod.com/2020/09/06/203/wordpress-cache-enabler-advanced-full-page-caching-guide/
2. Out of box defaults for Nginx, OpenLiteSpeed and Litespeed aren’t enough for very high concurrent traffic loads. From actual real world production from my high traffic paying consulting clients + synthetic benchmarks. Regardless at high traffic levels both Nginx and Litespeed/OpenLitespeed requiring further tuning beyond out of box installation anyway. High traffic is as in 500,000+ unique IP visitors/day and/or 15,000+ PHP related requests/second.
3. Comparing web servers, you left out how they respectively handle PHP requests. Nginx via PHP-FPM and OpenLiteSpeed and Litespeed via LSAPI PHP. Litespeed LSAPI is more closely integrated with Litespeed server hence partially better performance and definitely scales better than Nginx’s PHP-FPM. Nginx’s PHP-FPM is a separate server maintained and developed by PHP and not Nginx folks.
4. Litespeed’s LSAPI based PHP is definitely faster for non-cached dynamic php than PHP-FPM. uncached meaning without a cache layer in front of PHP which offloads PHP load elsewhere. OPcache is included in tests as it still part of PHP. When I mean non-cached I mean without varnish cache, cloudflare cache everything page rule, php-fpm fastgci_cache of dynamic PHP, litespeed cache of dynamic PHP or any other caching layer which moves the load away from PHP itself with exception of Opcache.
5. Litespeed isn’t the same as OpenLiteSpeed – some performance enhancements only are in Litespeed. And Litespeed 5.4+ is a totally different performance beast compared to either versions. I was surprised when I forked their benchmark tools to retest using their testing methodology. I wrote about it at https://www.webhostingtalk.com/showthread.php?t=1775139 along with doing a fork for early http2benchmark code. The forked http2benchmark code results are further down in that thread or at https://github.com/centminmod/http2benchmark/tree/extended-tests/examples which do show Litespeed 5.4 is faster than Litespeed 5.3 and Nginx official RPM installs configured more optimally by me (not Centmin Mod nginx still though so Nginx results there would be slower than Centmin Mod Nginx). Alot of my earlier tests of my own were with Litespeed 5.3 and lower and Centmin Mod Nginx which was alot closer and sometimes better for static files at least. But Litespeed 5.4 is new and faster than Litespeed 5.3. From https://www.webhostingtalk.com/showthread.php?t=1775139&p=10169960#post10169960
6. Well backend stack performance is important. Alot of load can sometimes be offloaded from the server to upstream i.e. Cloudflare CDN cache and other caching methods which can make the choice of web server less important sometimes. For example Cloudflare Automatic Signed Exchanges is a game changer especially if you get a lot of traffic from Google search https://blog.centminmod.com/2021/10/13/2511/testing-page-speed-with-cloudflare-automatic-signed-exchanges-google-search-cache/
Just my AUD$0.02 ^_^
Thanks for the extra insight and really nice to see you here, George. I’m a big fan and followed many of your threads over at WHT for years. 🙂
Thanks, Johnny! Definitely, your articles on server performance are what first caught my eye and I stayed for the WordPress side and info ^_^
What (managed) vps-type hosting would you recommend that uses LiteSpeed (preferably Enterprise), that’s not extremely expensive and has reasonable support?
There’s RunCloud. They only offer OpenLiteSpeed (not LiteSpeed Enterprise), but it’s plenty good enough. The price is cheap so no harm in trying.
That’s one of my options – however the security options (waf) are only in the top plan – which in total makes it not that very cheap (especially if you need it for one project…
I’ve read your recent post on CloudPage – however it’s a similar issue.
Looking into Rocket.net – not LiteSpeed based though, but have very good experiences from the past projects and it’s a complete solution, not a management panel…
If you understand what WAF really is….you’ll know it can be easily done in 2 seconds. Copy the 7G rules into your .htaccess and that’s it. 😉
Hi Johnny,
I have run several validation benchmarks using the K6 loader. The results can be found at https://www.webwhim.co.uk/how-to-select-a-wordpress-hosting-provider/#PHPBenchmarks . It has been my experience that the uncached performance of vanilla WordPress installations is independent of the software stack used (Nginx, Apache/Nginx, Litespeed) but is completely determined by CPU speed. The only exception is Raidboxes, which managed to destroy their uncached performance by editing core WordPress files.
A major selling point for end users is LiteSpeed’s Cache Engine plus its integration with Quic.cloud’s CDNs. This does not affect uncached performance, however.
Regards
Victor
No sense test.
The right test comparison possible is :
1) NGINX without Cache / OLS without Cache.
2) NGINX with Cache / OLS with LSCache
3) NGINX with Varnish Cache / OLS with LSCache
This is the combo on the real market. Any other test no make sense.
Just my two cent.
I forgot to mention who I am and what I do.
I am an expert, since 2005 I have been providing high performance hosting for WordPress sites that make up to 200 million page views per month and over 100,000 unique users connected per minute (Google Analytics in realtime).
We use NGINX and VARNISH based software stacks on the frontend and REDIS.IO / Memcached as page or object cache in the backend.
Objectively Open LiteSpeed ​​is not in the least taken into consideration as it is not possible to do complex operations, such as the split of the cache based on the device, the intelligent management of the webp delivery, the complex management of routing routing on other servers in different servers on the same subnet and many other things.
Not surprisingly, we in Italy as a company continue to take very demanding and very profitable companies as customers, whereas other competitors who have switched from Apache to Open LiteSpeed ​​because they are not expert enough to know how to install and configure NGINX as a reverse proxy and SSL terminator on Varnish. , they were forced to switch to OLS to fill their technical ignorance.
After all, you just need to do a search on similartech.com and compare the NGINX + Varnish and Open LiteSpeed ​​combos to make yourself practically NOBODY in the top500 sites uses it.
Enterprise level hosting (Pantheon.io, WP Engine, WordPress VIP, Fastly, Kinsta) use NGINX + VARNISH exclusively.
The reasons are very obvious.
I’m well aware of why large enterprise use NGINX and don’t have issues hiring dev-ops for complex NGINX configurations.
OLS however, delivers great results and more advanced features easier to the end user. End of story. (What you built on NGINX isn’t necessary or noticeably impactful for 99.9999% of people out there.)
Hi i also configure cloud infrastructure for popular sites, however i do not just do wordpress but websites where neither have any pre-set configurations or expertise for it. I use nginx because thats what i’m asked to use, and i have to use every trick just to get some of OLS performance. When doing massive testing, the biggest fail i find for both is simply IO, and i test on a massive scale (thousands/s) including of php. The biggest fail for both is php.
OLS/LSPHP – will handle all the requests i throw at it, but if i dont restrict memory for php (the RAM use restriction is a default config i just removed it to see what it could do), and database is backed up, it will end up hanging and using all the ram, and the server i tested on had 128GB of ram.
NGINX/PHP-fpm – will not handle all the request i throw at it. Some requests will fail when php instances have full CPU usage but it does not hang when the database is fully used. It eats ram over time till full then has to terminate hanged processes. In general php-fpm gives me more headaches than lsphp and hangs more with more issues.
Performance wise using a php website neither has any known expertise for, OLS/lsphp is way faster than nginx/php-fpm by a significant margin. For instance on the same server, if it can only handle 2000/s for nginx, OLS can handle up to 10000/s for a heavy php based site on that same server. I measure requirements not by visitors/month or users per month, but rather by requested actions/s as this is your limitation. You can easily handle a million users if they all take their turn, so thats why its better to use actions/s instead. I use the site’s official jmeter profile because it is generated by the website itself for testing, which simulates users doing something from logging in, browsing, posting,etc and i can tell you for sure you are going to be doing things not allowed in general in a company on nginx/php-fpm to even have a hint of performance that ols/lsphp can offer. Fully configured for both, ols/lsphp really wins when you have a content based php site.
The reason why companies use ls/lsphp is not because of lack of expertise, infact the reason why i’m not allowed to use litespeed is precisely the lack of expertise and bosses who are stuck with outdated means where i’m not allowed to use my own designs, and this includes many other components as well. I am always greeted with “we will not support you if you use litespeed” from vendors who have devops and are aws certified, meanwhile i can use any webserver needed. You need expertise to get the best out of either and you do have to configure litespeed, not just nginx. There was one provider i used before that had nginx + varnish caching. Not only was the service more expensive, but they charged for things that were free and offered a worse service compared to litespeed based hosts. On top of that the nginx/nginx + varnish cache kept interfering with content changes on the site whereas the really good thing about litespeed’s cache engine is that it can not only cache things on the webserver site to not touch the php, but it knows and can cache more things without stopping site updates from being shown unless you go crazy and start going in deep and manually touching settings.
A lot of internal politics is whats stopping litespeed from being popular, and from my own testing, even on the raw performance level, while litespeed itself is slower than nginx for static things (less ms to handle fewer static requests), but since most websites are dynamic and use php for instance, once taken into account, ls+lsphp is faster than nginx+php-fpm in response times, and litespeed is more flexible than nginx which is a reason why people use apache in the first place but litespeed still has far to go in load balancing than nginx as you can do complex configurations with both. You can have multiple dedicated lsphp servers to load balance but litespeed has less load balancing options than nginx.
To make a long story short: OLS is for non-technical (lazy) users.
Professional sysadmins appreciate the flexibility and power that Nginx+Varnish gives them. They never even consider OLS.
Personally I would never reach for proprietary software if I had an open-source alternative. Actually, here OLS is the alternative. All these benchmarks…. come on.
Hitch for SSL termination, Varnish at front-end and Nginx at back-end is at the speed of light.
Not to mention all the power and flexibility that BAN and PURGE give you in Varnish. Oh I forgot, OLS can some form of PURGE. 🙂
LOL. Well, there’s a clever way to turn a pro into a con. To make a long story short, some professional sysadmins feel awfully threatened by OLS, don’t they? 🙂
I think the resistance just comes down to familiarity with the web servers. Folks will always prefer what they’re familiar with.
I have around 12+yrs using Litespeed and 10+ yrs using Nginx and played with OpenLiteSpeed since it’s release including writing my own shell script auto installers for Litespeed standalone non-cPanel installs and OpenLiteSpeed auto installers with vhost creation routines so that I can change the default Litespeed and OpenLiteSpeed install configuration options.
I even built OpenLiteSpeed binaries that are faster than the one’s officially provided by Litespeed folks in their early YUM repos via https://github.com/litespeedtech/openlitespeed and https://openlitespeed.org/kb/install-openlitespeed-from-source-with-build-script/
Then, I got a question Sir George;
How to not install third-party tools or modules like ModSecurity, BoringSSL and Luajit during installation? Is there any way to customize the modules to build? Is there any option to prevent their installation?
for build.sh, I have no idea but for “configure”;
“./configure –with-brotli=no –with-iptogeo2=no –with-ip2loc=no –with-lua=no –with-bssl=no” didn’t work at all. It still tries to install it (on FreeBSD these addons create errors and the installation gets interrupted)
I’d like to achieve an OLS without those 3rd party addons, simply trying OpenLiteSpeed 1.7.14 with ./build.sh.
Thanks for share your knowledge, i am on a hosting which they use openlite i hase some problems whem i change permalinks on wordpress and the only solution theu give me is restart, do you know how to avoid this? I am looking to guidance for my new stack for my clients. Thanks you.
Ivan
no, to use either nginx or OLS you need skill, otherwise you wont be able to get the best of what they offer. Fully configured OLS is a lot faster and by a huge margin.
Varnish gave me more issues than litespeed’s cache engine, and ends up being slower as well. When using a provider that offered nginx and varnish it was not only more expensive and slower but you had to manually purge the cache on the provider just for any new content to show. You wont be able to host a forum on that. Meanwhile providers that used litespeed had less issues and the lscache engine worked flawlessly, so websites functioned fine while being faster.
Great and insightful post. I did some (load)-testing with Nginx and OLS as well on Runcloud, and in my setup Nginx seems to be faster than OLS, especially in uncached situations.
Did you by any chance watch the CPU and memory usage during the load-testing (both cached and uncached)?
From my personal point of view:
Arguing Nginx vs LiteSpeed is bonkers.
If youre using Nginx, thats fine.
If youre using LS, thats fine.
If youre using Apache, thats totally not fine.
And coming from Apache, making the jump to LS is way easier (htaccess ftw). Basically why i prefer LS over Nginx.
For me thats the most important thing. And there are so many shared hostings out there on Apache that could make a big improvement by jumping on LS with a greatly lower cost compared to jumping on Nginx.
Though, like the comment before mine, CPU and Mem usage wouldve been cool to see.
Cheers
How can compare anything what is dumb and can´t work with PHP. Nginx sending only html files…
I didn’t understand you. What are you trying to say?
He probably saying that nginx no any having lsphp/lsapi like litespeed. You be having to use php-fpm. Anyway, I be trying openlitespeed yesterday and configuration was being annoying. Also noticed later that it no any be supporting NSA selinux. But maybe I will trying again even if it no any supporting NSA selinux.
Yes, Nginx is for only static files good.
I try Nginx+Apache but vs OLS have no chance. Have on every 1 VPS two WordPress sites. WP All import pro plugin for nonstop updating .xml feeds.
With Hestia panel die in seconds, with Plesk few minutes-half hour before error and continoe with half speed. With aaPanel and OLS no problem ultra-heavy loads.
I don’t know what types of site You try, but for WP is the LiteSpeed best and the future with DDOS security.
Nginx+php-fpm be be working fine for my wordpress. Only recently some configuration problems I be having with geoip. But at least geoip be working out of box in nginx. Couldn’t get geoip working with openlitespeed yesterday when I be trying. Limiting access to wp-login.php by country code I was trying, but it was no any working. And then as I mentioning above, I also noticing that it no any be supporting NSA selinux unlike nginx either. Try google “openlitespeed selinux” and you will be seeing what I be saying.
Can someone, that is advanced in all this, please tell for sure if Nginx is better than OLS for static only websites?
I will not use PHP, just HTML pages.
Thank you!
For static HTML pages, it really doesn’t matter. As those are so easy to load anyway.
Also can I ask, is there any need to have some sort of server-caching in this case?
Or is it fine to just upload the static files?
I would like to use this ‘setup’:
– Cloudflare for DNS
– Bunny for CDN
– Publii for making the site files.
Thanks again.
Caching is for turning dynamic requests into static files. If your requests are already static, then no.
What i can say , on my end, I passed from LS to Rocket.net , and I am not an expert but without any configuration speed basically increase of 10 points in pagespeed . Probably the future is provide everything trough Cloudflare enterprise .
Will be more awesome to mix it with wp rocket or css js manager for wordpress 🙂
I think 90% of Litespeed is marketing. Speed difference is not really notably faster imo.