Just did a series of poor man's benchmarks against beta3, using Apache Benchmark with no concurrent requests (ab -n 100):
|
vanilla WP 1.5.1 |
3.90 req/s |
|
LP 1.1.0beta3 |
11.90 req/s |
|
LP 1.1.0beta3 with cache |
119.90 req/s |
I am too lazy to install wp-cache, but using Sencer's benchmark to do the math we should be about 50% faster than wp-cache. I will try to find the time to install wp-cache on this machine and have some real numbers, but not before we release beta3.
The speed with which you can serve a cache-hit is pretty much unrelated to how fast the blogsoftware itself is. So unless you found a totally new way of serving cached pages, I think it's unlikely the math will hold up. ;)
The blog engine has a direct influence on the speed of serving a cache hit. Working with static files, every line of code included/parsed/executed results in slower times: avoiding to include our main class and run a few lines of its constructor halved the cache serving time. An opcode cache will minimize this effect, of course.
My ex-partners at Blogo.it are converting their blogs from a pre-1.0 release to 1.1.0 with cache. The first benchmarks give a 20x increase in speed, and the bottleneck is their Celeron 900 HTTP server (which is doing quite a few other things besides serving their blogs). The load on the MySQL server is decreasing noticeably. More live stats in the next few days.