|
|
01-21-2011, 12:57 PM
|
|
Senior Member
|
|
Join Date: Dec 2004
Location: Brisbane, Australia
Posts: 142
|
|
Good news Litespeed 4.0.19 fixed the cache settings issue, now test.php gets cached while testnocache.php doesn't get cached with cache policy in 4.0.19 set as per below settings.
test.php header
Code:
view source
Content-Encoding gzip
Vary Accept-Encoding
Date Fri, 21 Jan 2011 20:54:52 GMT
Server LiteSpeed
Connection Keep-Alive
Keep-Alive timeout=5, max=100
X-LiteSpeed-Cache hit
Content-Length 79
X-Powered-By PHP/5.3.5
CurrentTime Fri, 21 Jan 2011 20:54:34 GMT
Content-Type text/html; charset=UTF-8
testnocache.php header
Code:
Content-Encoding gzip
Vary Accept-Encoding
Date Fri, 21 Jan 2011 20:54:59 GMT
Server LiteSpeed
Connection Keep-Alive
Keep-Alive timeout=5, max=100
X-Powered-By PHP/5.3.5
CurrentTime Fri, 21 Jan 2011 20:54:59 GMT
Content-Type text/html; charset=UTF-8
Content-Length 79
test.php
Code:
ab -k -n 1000 -c 200 192.168.56.101/test.php
This is ApacheBench, Version 2.0.40-dev <$Revision: 1.146 $> apache-2.0
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Copyright 2006 The Apache Software Foundation, http://www.apache.org/
Benchmarking 192.168.56.101 (be patient)
Completed 100 requests
Completed 200 requests
Completed 300 requests
Completed 400 requests
Completed 500 requests
Completed 600 requests
Completed 700 requests
Completed 800 requests
Completed 900 requests
Finished 1000 requests
Server Software: LiteSpeed
Server Hostname: 192.168.56.101
Server Port: 80
Document Path: /test.php
Document Length: 63 bytes
Concurrency Level: 200
Time taken for tests: 0.42640 seconds
Complete requests: 1000
Failed requests: 0
Write errors: 0
Keep-Alive requests: 1000
Total transferred: 346952 bytes
HTML transferred: 63000 bytes
Requests per second: 23452.16 [#/sec] (mean)
Time per request: 8.528 [ms] (mean)
Time per request: 0.043 [ms] (mean, across all concurrent requests)
Transfer rate: 7926.83 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 1.6 0 6
Processing: 0 2 1.0 3 5
Waiting: 0 2 1.0 3 5
Total: 0 3 1.5 3 11
Percentage of the requests served within a certain time (ms)
50% 3
66% 3
75% 3
80% 3
90% 6
95% 7
98% 8
99% 8
100% 11 (longest request)
testnocache.php
Code:
ab -k -n 1000 -c 200 192.168.56.101/testnocache.php
This is ApacheBench, Version 2.0.40-dev <$Revision: 1.146 $> apache-2.0
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Copyright 2006 The Apache Software Foundation, http://www.apache.org/
Benchmarking 192.168.56.101 (be patient)
Completed 100 requests
Completed 200 requests
Completed 300 requests
Completed 400 requests
Completed 500 requests
Completed 600 requests
Completed 700 requests
Completed 800 requests
Completed 900 requests
Finished 1000 requests
Server Software: LiteSpeed
Server Hostname: 192.168.56.101
Server Port: 80
Document Path: /testnocache.php
Document Length: 63 bytes
Concurrency Level: 200
Time taken for tests: 0.184258 seconds
Complete requests: 1000
Failed requests: 0
Write errors: 0
Keep-Alive requests: 1000
Total transferred: 323000 bytes
HTML transferred: 63000 bytes
Requests per second: 5427.17 [#/sec] (mean)
Time per request: 36.852 [ms] (mean)
Time per request: 0.184 [ms] (mean, across all concurrent requests)
Transfer rate: 1709.56 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 1.5 0 6
Processing: 20 28 14.6 21 89
Waiting: 20 28 14.4 21 88
Total: 20 29 15.7 21 89
Percentage of the requests served within a certain time (ms)
50% 21
66% 26
75% 27
80% 28
90% 59
95% 67
98% 72
99% 83
100% 89 (longest request)
The test.php and testnocache.php file used contains
Code:
<?php
header('CurrentTime: '.gmdate('D, d M Y H:i:s', time()).' GMT',true);
echo "time()=" . time() . "<br>";
echo "date()=" . date(DATE_RFC822) ."<br>";
?>
Apache httpd.conf contains
Code:
#disk cache
<IfModule mod_cache.c>
<IfModule mod_disk_cache.c>
CacheRoot /lscache/
#CacheEnable disk /
</IfModule>
</IfModule>
Litespeed 4.0.18 cache policy is set to
Code:
Enable Cache:No
Cache Request with Query String:Yes
Cache Request with Cookie:Yes
Cache Response with Cookie:Yes
Ignore Request Cache-Control:Yes
Ignore Response Cache-Control:Yes
|
01-21-2011, 01:18 PM
|
|
LiteSpeed Staff
|
|
Join Date: May 2003
Location: New Jersey
Posts: 7,590
|
|
|
It may have something to do with how "ab" works, it send more than 1000 requests during the test, and close all connection once number of responses reaches 1000.
It is better to do one test run of
ab -k -n 5000 -c 200 ...
than running ab 5 times with "-n 1000".
In theory, the result should be same, but it is not due to the way "ab" works.
|
01-21-2011, 01:25 PM
|
|
Senior Member
|
|
Join Date: Dec 2004
Location: Brisbane, Australia
Posts: 142
|
|
i see, maybe httperf would be better ?
i.e.
Quote:
|
httperf --server 192.168.56.101 --uri /test.php --num-conn 1000 --num-call 1 --rate 200 --timeout 5
|
All above results are without any php caching. So maybe also try with xcache and eaccelerator ? Part 3 
|
01-21-2011, 01:51 PM
|
|
LiteSpeed Staff
|
|
Join Date: May 2003
Location: New Jersey
Posts: 7,590
|
|
yes, httperf is better in this regard. However, httperf does rate limiting, not like "ab" sending request as fast as possible, so, httperf may report lower req/s than "ab" if the rate limit cannot push the server to 100%.
Opcode cache will help for all web servers.
There are other respects need to be taken into consideration other than the raw req/s, like server load, memory usage, CPU utilization, etc. 
|
01-21-2011, 01:58 PM
|
|
Senior Member
|
|
Join Date: Dec 2004
Location: Brisbane, Australia
Posts: 142
|
|
Yeah apache at >600 concurrency levels, the cpu utilization is in double digits at end of 5th ab run while litespeed hasn't budged much cpu load wise
Maybe you need to come up with your own benchmark script called litespeedbench Could output cpu utilisation/load + mem usage as well as rps figures.
Last edited by eva2000; 01-21-2011 at 02:02 PM..
|
01-22-2011, 09:19 AM
|
|
Senior Member
|
|
Join Date: Dec 2004
Location: Brisbane, Australia
Posts: 142
|
|
Not sure if this is directly related, but i also installed nginx 0.8.54 /php-fpm 5.3.5 for comparisons, and it's even worse with nginx, test.php apachebench run with =>200 concurrency, 2nd run already drops from 5500 rps to consistent 330 rps.
Looked in my messages log and found heaps of logged entries for
Code:
tail -500 /var/log/messages
Jan 23 02:18:22 localhost kernel: printk: 206 messages suppressed.
Jan 23 02:18:22 localhost kernel: ip_conntrack: table full, dropping packet.
Code:
cat /proc/sys/net/ipv4/ip_conntrack_max
32760
so raised value in /etc/sysctl.conf
Code:
#net.ipv4.ip_conntrack_max=32760
net.ipv4.ip_conntrack_max=262144
no more ip_conntrack messages, but still subsequent apachebench runs back to back end with lower requests per second with test.php test, worse being nginx (can't handle =>200 concurrency), then apache (past >600 concurrency), while litespeed handles it better (up to 1000 concurrency).
maybe need some tuning http://timanovsky.wordpress.com/2009...-ip_conntrack/
To confirm the issue ran some static test.txt file tests and nginx was fine, while apache still exhibited the problem at higher concurrency levels http://www.litespeedtech.com/support...ead.php?t=4617
Edit: solved nginx php issue changed php-fpm from tcp to using unix socket
Last edited by eva2000; 01-24-2011 at 12:18 PM..
|
01-22-2011, 12:30 PM
|
|
LiteSpeed Staff
|
|
Join Date: May 2003
Location: New Jersey
Posts: 7,590
|
|
|
One of the factor is what "ab" did at the end of the benchmark tests. It can affect one web server more than to the others. To reduce the impact of this, just combine multiple runs into one large run.
Another factor is the way web server works, perfork, threaded or event-driven. perfork has poor scalability is for sure.
|
01-23-2011, 04:38 AM
|
|
Senior Member
|
|
Join Date: Dec 2004
Location: Brisbane, Australia
Posts: 142
|
|
When you mean multiple runs you mean
2 runs
ab & ab
4 runs
ab & ab & ab & ab
?
Curious as to performance of my virtualbox guest server, so ran Unixbench 5.1.2 as per http://www.webhostingtalk.com/showthread.php?t=924581 to see how it compares.
- VirtualBox CentOS 5.5 64bit Guest
- Xeon W3540 @3408Mhz (Assigned 2 cores only) - linux reports @3375Mhz
- 64MB allocated GPU memory
- 1GB out of 6GB DDR3-1550Mhz
- 20GB out of 640GB Samsung SATAII
Code:
========================================================================
BYTE UNIX Benchmarks (Version 5.1.2)
System: localhost.localdomain: GNU/Linux
OS: GNU/Linux -- 2.6.18-194.32.1.el5 -- #1 SMP Wed Jan 5 17:52:25 EST 2011
Machine: x86_64 (x86_64)
Language: en_US.utf8 (charmap="UTF-8", collate="UTF-8")
CPU 0: Intel(R) Xeon(R) CPU W3540 @ 2.93GHz (6750.1 bogomips)
Hyper-Threading, x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET
CPU 1: Intel(R) Xeon(R) CPU W3540 @ 2.93GHz (6733.8 bogomips)
Hyper-Threading, x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET
21:45:27 up 1:11, 2 users, load average: 0.00, 0.01, 0.00; runlevel 5
------------------------------------------------------------------------
Benchmark Run: Sun Jan 23 2011 21:45:27 - 22:09:32
2 CPUs in system; running 1 parallel copy of tests
Dhrystone 2 using register variables 19101079.5 lps (10.0 s, 7 samples)
Double-Precision Whetstone 3771.0 MWIPS (10.0 s, 7 samples)
Execl Throughput 4394.2 lps (30.0 s, 2 samples)
Pipe Throughput 1457951.0 lps (10.0 s, 7 samples)
Pipe-based Context Switching 29373.4 lps (10.0 s, 7 samples)
Process Creation 13113.2 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 7997.6 lpm (60.0 s, 2 samples)
Shell Scripts (16 concurrent) 901.5 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 1756.9 lpm (60.0 s, 2 samples)
System Call Overhead 1211587.2 lps (10.0 s, 7 samples)
System Benchmarks Partial Index BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 19101079.5 1636.8
Double-Precision Whetstone 55.0 3771.0 685.6
Execl Throughput 43.0 4394.2 1021.9
Pipe Throughput 12440.0 1457951.0 1172.0
Pipe-based Context Switching 4000.0 29373.4 73.4
Process Creation 126.0 13113.2 1040.7
Shell Scripts (1 concurrent) 42.4 7997.6 1886.2
Shell Scripts (16 concurrent) --- 901.5 ---
Shell Scripts (8 concurrent) 6.0 1756.9 2928.1
System Call Overhead 15000.0 1211587.2 807.7
========
System Benchmarks Index Score (Partial Only) 916.9
------------------------------------------------------------------------
Benchmark Run: Sun Jan 23 2011 22:09:32 - 22:33:38
2 CPUs in system; running 2 parallel copies of tests
Dhrystone 2 using register variables 37996013.0 lps (10.0 s, 7 samples)
Double-Precision Whetstone 7678.6 MWIPS (9.7 s, 7 samples)
Execl Throughput 8655.7 lps (29.9 s, 2 samples)
Pipe Throughput 2863704.1 lps (10.0 s, 7 samples)
Pipe-based Context Switching 739539.5 lps (10.0 s, 7 samples)
Process Creation 25690.5 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 14271.1 lpm (60.0 s, 2 samples)
Shell Scripts (16 concurrent) 1001.8 lpm (60.1 s, 2 samples)
Shell Scripts (8 concurrent) 1990.0 lpm (60.0 s, 2 samples)
System Call Overhead 2279039.9 lps (10.0 s, 7 samples)
System Benchmarks Partial Index BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 37996013.0 3255.9
Double-Precision Whetstone 55.0 7678.6 1396.1
Execl Throughput 43.0 8655.7 2012.9
Pipe Throughput 12440.0 2863704.1 2302.0
Pipe-based Context Switching 4000.0 739539.5 1848.8
Process Creation 126.0 25690.5 2038.9
Shell Scripts (1 concurrent) 42.4 14271.1 3365.8
Shell Scripts (16 concurrent) --- 1001.8 ---
Shell Scripts (8 concurrent) 6.0 1990.0 3316.7
System Call Overhead 15000.0 2279039.9 1519.4
========
System Benchmarks Index Score (Partial Only) 2226.9
top stats straight after unixbench run
Code:
top - 22:34:03 up 2:00, 2 users, load average: 6.78, 7.13, 4.75
Tasks: 124 total, 1 running, 123 sleeping, 0 stopped, 0 zombie
Cpu(s): 14.4%us, 13.9%sy, 0.0%ni, 70.8%id, 0.4%wa, 0.0%hi, 0.4%si, 0.0%st
Mem: 1026824k total, 688652k used, 338172k free, 31276k buffers
Swap: 2064376k total, 0k used, 2064376k free, 457076k cached
Code:
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 26
model name : Intel(R) Xeon(R) CPU W3540 @ 2.93GHz
stepping : 5
cpu MHz : 3375.078
cache size : 6144 KB
physical id : 0
siblings : 2
core id : 0
cpu cores : 2
apicid : 0
fpu : yes
fpu_exception : yes
cpuid level : 5
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx lm constant_tsc pni ssse3 lahf_lm
bogomips : 6750.15
clflush size : 64
cache_alignment : 64
address sizes : 36 bits physical, 48 bits virtual
power management:
processor : 1
vendor_id : GenuineIntel
cpu family : 6
model : 26
model name : Intel(R) Xeon(R) CPU W3540 @ 2.93GHz
stepping : 5
cpu MHz : 3375.078
cache size : 6144 KB
physical id : 0
siblings : 2
core id : 0
cpu cores : 2
apicid : 0
fpu : yes
fpu_exception : yes
cpuid level : 5
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx lm constant_tsc pni ssse3 lahf_lm
bogomips : 6733.77
clflush size : 64
cache_alignment : 64
address sizes : 36 bits physical, 48 bits virtual
power management:
Last edited by eva2000; 01-23-2011 at 04:41 AM..
|
01-24-2011, 09:16 AM
|
|
LiteSpeed Staff
|
|
Join Date: May 2003
Location: New Jersey
Posts: 7,590
|
|
|
I mean if you run "ab" 5 times with -n 1000, it will be more accurate result if you do "ab" once with -n 5000.
"ab" will send maybe around 2000 requests when you do " -n 1000 -c 1000", after it receives the first 1000 response, it just close all connection, however, on web server side, those extra requests are still being processed, that's why the first "ab" run get better result.
|
01-24-2011, 10:46 AM
|
|
Senior Member
|
|
Join Date: Dec 2004
Location: Brisbane, Australia
Posts: 142
|
|
I see. Will test with higher requests rate in future. I'll install test vB 3.8.6 pl1 forum now for more real life testing
|
| Thread Tools |
|
|
| Display Modes |
 Linear Mode
|
Posting Rules
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
HTML code is Off
|
|
|
All times are GMT -7. The time now is 09:20 PM.
|
|
