I am a moron.

This is embarassing, you'd think that 'siege
reported the same results...doesn't use
keepalive' would have lit a bulb somewhere.

Setting a valid Content-Length gives
a much nicer picture, and it still only uses
half the cpu:

Server Software: nginx
Server Hostname: amaterasu
Server Port: 443
SSL/TLS Protocol: TLSv1/SSLv3,RC4-SHA,2048,128

Document Path: /test.php
Document Length: 13 bytes

Concurrency Level: 900
Time taken for tests: 13.894 seconds
Complete requests: 90000
Failed requests: 5
(Connect: 0, Receive: 0, Length: 5, Exceptions: 0)
Write errors: 0
Keep-Alive requests: 89372
Total transferred: 29680613 bytes
HTML transferred: 1169948 bytes
Requests per second: 6477.66 [#/sec] (mean)
Time per request: 138.939 [ms] (mean)
Time per request: 0.154 [ms] (mean, across all concurrent requests)
Transfer rate: 2086.16 [Kbytes/sec] received

Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 15 146.6 0 2346
Processing: 62 123 110.9 102 3125
Waiting: 0 123 108.5 102 1679
Total: 62 138 237.3 102 3125

Percentage of the requests served within a certain time (ms)
50% 102
66% 113
75% 122
80% 129
90% 155
95% 193
98% 413
99% 1204
100% 3125 (longest request)

----

Thank you for pointing out my stupidity, I'll go hide
in shame now. : )

Max Wrote:
-------------------------------------------------------
> The robots.txt via SSL test shows 89708 of 90000
> Keep-Alive requests, which means the same SSL
> connection was reused 99.67 % of the time, which
> is why you got such good test results.
>
> But the test.php via SSL test shows 0 out of 9000
> Keep-Alive requests, which means that a new SSL
> connection was established for each and every
> successful request. Considering how CPU intensive
> SSL is, your test results are really not that bad.
>
> Max
> _______________________________________________
> nginx mailing list
> nginx@nginx.org
> http://mailman.nginx.org/mailman/listinfo/nginx