Sorry Wayne. Simple load test of Crazyontap. (1 request = 1s)
As I was testing my app, thought I might test CrazyOnTap. Here are the simple results. Basically, I have an application that requests a page from the website and downloads it.
My site (main page with dynamic content)
My site locally from a really fast machine.
July 3rd, 2007 1:23pm
July 3rd, 2007 1:29pm
Google caches. Its first Bush took longer to grab than the rest.
It's be more interesting throwing a pair of randomly generated words at it, 500 times.
July 3rd, 2007 1:56pm
I hope wayne doesnt call my company (cough, cough) because I am doing all these requests to CoT.
July 3rd, 2007 2:21pm
A couple of things about CoT:
* There's no HTTP caching on CoT at all (and it's a prime candidate for it given it's simplicity).
* There's no server-side caching on CoT -- every request hits the database (and again, since everybody sees the same view, it would benefit from that greatly).
* CoT is on a shared server with other high traffic database-intensive sites.
But the performance of CoT is pretty reasonable -- you don't often have to wait so no effort has been put into speeding things up. Server-side caching of rendered versions of the pages would make the biggest difference. Maybe in the next version of FruitShow I'll add something like that. I have a lot of experience adding all kinds of caching and performance enhancements to sites.
July 3rd, 2007 2:32pm
"I have a lot of experience adding all kinds of caching and performance enhancements to sites."
Interesting. Is that done on the php side or through apache or none of the above.
July 3rd, 2007 2:34pm
PHP side, Apache site, and MySQL side.. This server is not as tweaked out as any of the other servers. I should actually go check the configuration -- I spend so much more time dealing with the other, overloaded, servers that this server might not be running optimally.
This server does not have PHP code caching (maybe I'll add it) so on each request the PHP code is read from disk, compiled, and executed. On one my other servers, I have PHP configured to run entirely from memory -- it doesn't hit the disk at all. Even the statically cached elements are stored in memory.
The PHP side gives you the biggest bang for you buck -- store rendered pages or elements on disk (or memory) and then read them back up. My database code is designed to lazy-connect, so it doesn't make the connection until I actually make a query. On a few sites, pages are built up from different elements all cached at different rates (1 hour for article area, 5 minutes for recent post list, etc). CoffeeGeek.com used to have a bit of layered caching: pages were cached fully rendered and then user-specific customizations were applied to the cached version before rendering it out.
July 3rd, 2007 2:46pm
> Its first Bush took longer to grab than the rest.
Yeah, I remember that feeling from puberty.
man on the stair
July 4th, 2007 11:10am
thanks, man on stair, someone got it.
July 5th, 2007 12:31am