All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
New LiteSpeed (Open and Enterprise) vs. nginx vs. Apache Benchmarks!
Howdy all,
Been a while since I posted here. Figured some of y'all might be interested. We (LiteSpeed Technologies) just posted some new benchmarks:
summary blog post
small static file benchmarks
small static file HTTPS benchmarks
These are just the first in a series of benchmarks we're planning. We're starting with small static files and working our way through dynamic content up to web applications. (This may take a while.)
In these benchmarks, LiteSpeed and OpenLiteSpeed dusted the competition. I'm sure some of you are going to take issue with that. That's kind of what I'm hoping. Y'all are some opinionated people and I'd love to hear your feedback on the tests — either here, on our forum, on the blog, through email, whatever. We're perfectly happy to retest if someone finds something wrong.
Looking forward to it.
m
- Are you surprised by the benchmark results?51 votes
- Yes! I knew LiteSpeed was amazing, but... Wow!25.49%
- Yes! You totally screwed it up. Here's what you should have done...11.76%
- No! I already knew LiteSpeed was the best!  9.80%
- No! Of course LiteSpeed won LiteSpeed's benchmarks... Let's see some that aren't fixed.19.61%
- Pfffft... Who cares about benchmarks?33.33%
Comments
how dare you to insult the overlord nginx
Meh. I ain't scared of him.
Now post benchmarks showing how fast each can serve up galleries of high-resolution porn.
Just waiting for htaccess in OLS..
Based on these benchmarks, the question why LSWS and not OLS arises?
Honestly, there's a couple major flaws with the testing procedure.
First of all, LiteSpeed is a web server daemon targeted towards high-end traffic levels. High-end traffic levels wouldn't use OpenVZ, wouldn't use only 8gb of ram, and wouldn't be gimped to half a CPU core. This in mind, I'd like to see a benchmark with a fully capable machine.
Second of all, if you're worried about AB not being able to produce enough load, watch the client's CPU usage.
All in all, I think the tests do hold some water. However, what needs to be considered more is how much each web server performs under more performant systems, including an up-to-date kernel and operating system with hardware that is not purposefully gimped. I think that's where you'll see which systems are more or less performant under what workloads, and which ones produce the most bang for their buck.
I agree with everything Rallias said - there's definitely a few flaws in the testing procedure.
The target audience (here) overwhelmingly does, though
The test was done by the company selling the software, conflict of interest.
Mun
Well, I haven't benchmarked Litespeed at all, but I got completely different results.
Apache with mpm_event had about 16 000 requests per second and Nginx up to 35 000 and the server was only on a VM...
I'm not sure what you exactly have done guys, but I really got different results (also on the PHP benchmark)
Nothing to say. Just wanted to see this again.
Sorry for disappearing for so long. I'll stry to answer these kind of in order.
We're considering some form of .htaccess compatibility for OpenLiteSpeed. It won't be full .htaccess compatibility like Enterprise, but we want to make OpenLiteSpeed easier to use. (Just not quite as easy as Enterprise.) This will take a while (we've got a lot on our plate), but one of the leading ideas is a function that you could run to red .htaccess files and integrate them in your setting on restart. Basically this means you can update your .htaccess files every time you upgrade your applications, but you can't do shared hosting.
Well, as pointed out by ihatetonyy, OpenLiteSpeed doesn't have .htaccess compatibility. It's also not compatible with hosting control panels. It runs really fast, supports a lot of features, can understand Apache rewrite syntax, and has a nifty GUI, but it's not a drop-in Apache replacement. You have to buy a license to get a drop-in replacement.
m
Exactly! @Rallias @manacit , we sell a lot of VPS and Ultra VPS licenses. A lot a lot. And those licenses are limited to 2GB and 8GB of RAM respectively. LSWS is all about getting more out of your hardware. Instead of upgrading their little VPS, people put LiteSpeed on it and save money.
Aren't you at work? It's bad to watch porn while you're at work.
Well... I never!
I'm shocked and disheartened that you would think such things! I know I picked "No! I already knew LiteSpeed was the best!" in the poll above.
Link to your configs. We'd love to go over them. Ours are available here: http://www.litespeedtech.com/packages/benchmark/configs.tgz
http://turnkeye.com/blog/nginx-vs-litespeed-test-magento/
interesting info. seems litespeed didn't outperform nginx.
I see no reason to try OpenLiteSpeed after this.
A lot can change in 3 years
good point flip.
Just converted a customer to LiteSpeed the other day. cPanel server with one high-traffic site (around 2M pageviews per day) and ~50 low traffic sites.
Apache mod_fcgid, mpm prefork + nginxcp vs LiteSpeed:
LiteSpeed rulez!
any chance of offering discounts for Litespeed VPS licenses? I always wanted to get one
Litespeed price lill more high. Hope so they give us some discount.
George_Fusioned
So you did not use FPC at all on nginx? And started to use it on LS?
Sorry, guys. No discounts (except for bulk purchases/reselling). Generally, we figure if you need LSWS, you're saving money (on hardware, optimization, troubleshooting) by buying it.
Who was it that said "Free is the most expensive price"?
m
Said thank you on vpsBoard, but might as well repeat it here. Thanks!
Benchmarks and speed aside, this decrease in load, this smooth handling of high traffic and spikes, is the real point of LiteSpeed.
m
Seen this benchmark before. I could understand if the results were similar, but the huge difference makes me think something is fishy. Of course, I guess it's a little late to ask to see their configs... I'd love to see someone reproduce those results, but that's probably asking too much. Benchmarks are a pain.
m
Not really sure what you mean. Maybe there was a misunderstanding? You have to buy a license to get Enterprise edition (which has the features we were talking about). OpenLiteSpeed, of course, does not require a license.
m
You are not giving all of your hard work away for free. Therefore, your company is evil for trying to make money and may you all burn in hell. Because everything should be free, and unicorns should crap candy. Ironically, we will go about our day jobs where we get paid to do work.