litespeed on High I/O amazon ec2

[gdewey]

can someone make a guess how many concurrent users can LITESPEED handle on a site with this setting (considering all content is static and the rest in cache) ?.. all the dynamic content comes from a proxy context via xmlhttprequest (ajax) to a xml cached content. the proxy-target server is other than the lsws

High I/O Quadruple Extra Large Instance (hi1.4xlarge)

60.5 GiB of memory
35 EC2 Compute Units (16 virtual cores*)
2 SSD-based volumes each with 1024 GB of instance storage
64-bit platform
I/O Performance: Very High (10 Gigabit Ethernet)
Storage I/O Performance: Very High***
EBS-Optimized Available: No**
API name: hi1.4xlarge

*8 cores + 8 hyperthreads for 16 virtual cores

I am trying to prepare for an event we have every year that has over 90,000 concurrent connections for 24 hrs. We normally loadbalance 2 or 3 cache servers. But I want to give litespeed a chance for this year's.

And yes this is considering the Enterprise 8 CPU version.

 

[NiteWave]

the 90k connections should be no problem for litespeed to handle, but it may hit some hardware / OS limit for example NIC throughput.

assume all 90K connections are downloading large videos, for 10Gbit NIC, each connection will get max 11 bytes/second speed ? this speed won't be able to view video smoothly.

anyway, 90k connections looks a huge number. maybe have to prepare a few servers to serve smoothly.

 

[gdewey]

cache in front of webserver

tks for your reply. one question related to this

does it make sense to place a cache server infront of lsws that serves only static conntent?

 

[NiteWave]

no need at all, it only adds extra load.

only dynamic page need cache; static page should be served by litespeed directly. you can do benchmark testing, see if litespeed will outperform a cache server when serve static pages. when request a static file many times, linux will put the file in cache automatically, so litespeed will serve the same request more quickly.

 

[gdewey]

reply

well the only advantages I was thinking about the cache server where

1. that site is on ram of the cache server (that can also be done in here as you mention)
2. if origin server is down the cache server can serve older content until the origin is back.
3. also it has some dos protection, but so as lsws

let me make a benchmark between the 2. I just got the LOAD UI application to virtualize users for soap and http request

 

[gdewey]

speed

it seems like we are getting better benchmark WITH ls using using the exact same content ON

LITESPEED ENTEPRISE 2 CORE on a dedicated AMAZON WS medium instance

VS

CLOUDFLARE Business PLAN FrontEnd

VS

AICACHE Cache Server on AMAZON WS medium instance

is there a way to auto minify html on LS?

 

[NiteWave]

auto minify html should not be done by a web server.
the standard way is to gzip the html, which reduce the size and not change the original content.