Task / Discussion

This is the discussion related to the wiki page Task.

page 1 of 212next »

michal frackowiak 02 Oct 2009 07:36

Here is a small hint for developers:

The displayed result does not need to be shockingly up-to-date. The information listed on the main page (http://www.wikidot.com) is also cached for some period of time.

It should be more important to create an efficient way of fetching and displaying data and handling dozens of req/s than displaying up-to-date content. This is why we encourage using Memcached.

---

Michał Frąckowiak @ Wikidot Inc.
Visit my blog at michalf.me

Last edited on 02 Oct 2009 08:14 by michal frackowiak + Show more

plisken 04 Oct 2009 10:47

Just of the curiosity — this "Featured Sites" section is where exactly?

Gabrys 04 Oct 2009 11:48

@plisken: Featured Sites section is on the very center of http://www.wikidot.com/

---

Piotr Gabryjeluk
visit my blog

Benchmark ;>

TeRq 04 Oct 2009 20:03

Hello,

testing environment:
CPU: AMD Athlon 2000+
RAM: 768MB DDR1
HDD: 40GB Seagate Baracuda

My test :D

+ 10000/100 (starting with clean cache) Memcached

terq@kotki:~$ ab -n 10000 -c 100 http://localhost/wikidot
This is ApacheBench, Version 2.3 <$Revision: 655654 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking localhost (be patient)
Completed 1000 requests
Completed 2000 requests
Completed 3000 requests
Completed 4000 requests
Completed 5000 requests
Completed 6000 requests
Completed 7000 requests
Completed 8000 requests
Completed 9000 requests
Completed 10000 requests
Finished 10000 requests

Server Software:        Apache/2.2.11
Server Hostname:        localhost
Server Port:            80

Document Path:          /wikidot
Document Length:        3879 bytes

Concurrency Level:      100
Time taken for tests:   120.146 seconds
Complete requests:      10000
Failed requests:        0
Write errors:           0
Total transferred:      41762240 bytes
HTML transferred:       38941281 bytes
Requests per second:    83.23 [#/sec] (mean)
Time per request:       1201.461 [ms] (mean)
Time per request:       12.015 [ms] (mean, across all concurrent requests)
Transfer rate:          339.45 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0  516 204.7    519    1136
Processing:    25  680 515.2    614    6593
Waiting:        5  612 519.2    522    6593
Total:        607 1196 433.7   1139    6747

Percentage of the requests served within a certain time (ms)
  50%   1139
  66%   1145
  75%   1154
  80%   1168
  90%   1326
  95%   1410
  98%   1534
  99%   4426
 100%   6747 (longest request)

Cant make 10000/1000… its kiling my PC :>

Now writeing readme, then pack everything and sending it to wikidot.com

gl ALL

---

Bartłomiej Bąkowski, Wikidot Inc.
';.;' TeRq

Last edited on 06 Oct 2009 11:46 by TeRq + Show more

michal frackowiak 05 Oct 2009 07:33

A small hint is that if

1. the data is already in the cache,
2. you optimize the part of your script that displays the data

you could get a few hundreds of req/s quite easily. But performance aside, we really look much more at overall design (even with such a small task) and elegance of code, which reflects your coding habits.

---

Michał Frąckowiak @ Wikidot Inc.
Visit my blog at michalf.me

TeRq 05 Oct 2009 08:24

I‘m not sure, is that hint for me…
if so, the higher performance on my PC its a some kind of abstraction :>
ma machine handle ~400req/s on
<?php echo ’Hello world.'; ?>
script ;)
so 80req/s its an magnificent result
Of course in my opinion.

Yes, I could throw out OO api (from main page cache) and use some inline scripting but is it wourth ??

regards

PS. Script was already send to moc.todikiw|rtoip#moc.todikiw|rtoip

---

Bartłomiej Bąkowski, Wikidot Inc.
';.;' TeRq

michal frackowiak 05 Oct 2009 09:06

Well, not necessarily to anyone particular, but we have solutions that go down below 1 req/s. Honestly I have not seen your solution yet (Piotr did not share it, still waiting), but 80 req/s is reasonable, especially if you machine can do 400 max.

And we have also written our own solutions to this task and we are getting ~ 800 req/s when data is in the cache, on a dual-core 2.6GHz machine. But there are many external factors, hardware and software.

We will go through it, probably together with authors of solutions. No worries.

BTW: are you using APC or eAccelerator?

---

Michał Frąckowiak @ Wikidot Inc.
Visit my blog at michalf.me

TeRq 05 Oct 2009 09:22

@michal frackowiak

BTW: are you using APC or eAccelerator?

no i dont… (as I know)

could u past me benchmark from U testing Environment 10000/100
I`m so curious

---

Bartłomiej Bąkowski, Wikidot Inc.
';.;' TeRq

michal frackowiak 05 Oct 2009 09:32

ab -n 10000 -c 100 http://127.0.0.4:8081/task.php

+ Show the result

This is ApacheBench, Version 2.3 <$Revision: 655654 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking 127.0.0.1 (be patient)
Completed 1000 requests
Completed 2000 requests
Completed 3000 requests
Completed 4000 requests
Completed 5000 requests
Completed 6000 requests
Completed 7000 requests
Completed 8000 requests
Completed 9000 requests
Completed 10000 requests
Finished 10000 requests

Server Software:        lighttpd/1.4.22
Server Hostname:        127.0.0.4
Server Port:            8081

Document Path:          /task.php
Document Length:        2861 bytes

Concurrency Level:      100
Time taken for tests:   12.056 seconds
Complete requests:      10000
Failed requests:        0
Write errors:           0
Total transferred:      29920000 bytes
HTML transferred:       28610000 bytes
Requests per second:    829.44 [#/sec] (mean)
Time per request:       120.563 [ms] (mean)
Time per request:       1.206 [ms] (mean, across all concurrent requests)
Transfer rate:          2423.52 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    1   0.9      0      21
Processing:    31  119  15.5    119     217
Waiting:       28  116  15.2    115     211
Total:         34  120  15.4    120     217
WARNING: The median and mean for the initial connection time are not within a normal deviation
        These results are probably not that reliable.

Percentage of the requests served within a certain time (ms)
  50%    120
  66%    123
  75%    125
  80%    127
  90%    133
  95%    142
  98%    161
  99%    196
 100%    217 (longest request)

This is on a 64-bit Mac. FastCGI and eAccelerator might be adding some extra boost here, but this setup is new so I did not bother to much tuning. The real application have other bottlenecks.

---

Michał Frąckowiak @ Wikidot Inc.
Visit my blog at michalf.me

TeRq 05 Oct 2009 09:41

@michal frackowiak
thx,
but I thought about my script on U testing Environment ;) [starting with clean cache]
(test was mailed from: lp.qret|tkatnok#lp.qret|tkatnok)

---

Bartłomiej Bąkowski, Wikidot Inc.
';.;' TeRq

page 1 of 212next »