Performance comparison 1.2.5 to .1.4.1. - Update

[Trip]

Ok, after I read some people complaining about performance I wanted to proof them false and made some benchmarks on my local machine.
The results were not expected, not to say remarkable.
I used apache stress tool on a local machine which is not very realistic but for my opinion sufficient to see some tendencies.
I used the standard theme and a categorie page with 10 products.
If you want to test it own your own you can use

ab -kc 10 -t 30 http://yourshop

on your linux box.
This means your server is hammered with 10 concurrent user for 30 seconds


Prestashop 1.2.5 (worst) Page Speed Score 77/100

Concurrency Level:      10
Time taken for tests:   30.189 seconds
Complete requests:      460
Failed requests:        59
  (Connect: 0, Receive: 0, Length: 59, Exceptions: 0)
Write errors:           0
Keep-Alive requests:    0
Total transferred:      21635048 bytes
HTML transferred:       21296396 bytes
Requests per second:    15.24 [#/sec] (mean)
Time per request:       656.274 [ms] (mean)
Time per request:       65.627 [ms] (mean, across all concurrent requests)
Transfer rate:          699.87 [Kbytes/sec] received

Connection Times (ms)
             min  mean[+/-sd] median   max
Connect:        0    0   0.9      0       7
Processing:   333  647 109.9    649     938
Waiting:      330  642 108.7    643     917
Total:        333  647 109.9    649     938

Percentage of the requests served within a certain time (ms)
 50%    649
 66%    682
 75%    726
 80%    753
 90%    783
 95%    809
 98%    891
 99%    907
100%    938 (longest request)

Prestashop 1.4.1 Smarty 3 without optimisations
Page Speed Score: 80/100

only completes 236 request (460 PS 1.2.5) which means that it takes double the time to serve a page.

Concurrency Level:      10
Time taken for tests:   30.581 seconds
Complete requests:      236
Failed requests:        0
Write errors:           0
Keep-Alive requests:    0
Total transferred:      11561054 bytes
HTML transferred:       11334372 bytes
Requests per second:    7.72 [#/sec] (mean)
Time per request:       1295.793 [ms] (mean)
Time per request:       129.579 [ms] (mean, across all concurrent requests)
Transfer rate:          369.19 [Kbytes/sec] received

Connection Times (ms)
             min  mean[+/-sd] median   max
Connect:        0    0   1.2      0       7
Processing:   737 1263 247.8   1267    1739
Waiting:      727 1247 248.0   1255    1707
Total:        737 1264 247.9   1267    1739

Percentage of the requests served within a certain time (ms)
 50%   1267
 66%   1457
 75%   1502
 80%   1511
 90%   1543
 95%   1632
 98%   1677
 99%   1704
100%   1739 (longest request)

Prestashop 1.4.1 with all optimisations (except MCashe and Media Server)
Page Speed Score: 88/100 but has the worst real performance regarding the server load.

Concurrency Level:      10
Time taken for tests:   30.088 seconds
Complete requests:      159
Failed requests:        0
Write errors:           0
Keep-Alive requests:    0
Total transferred:      6487872 bytes
HTML transferred:       6335673 bytes
Requests per second:    5.28 [#/sec] (mean)
Time per request:       1892.340 [ms] (mean)
Time per request:       189.234 [ms] (mean, across all concurrent requests)
Transfer rate:          210.57 [Kbytes/sec] received

Connection Times (ms)
             min  mean[+/-sd] median   max
Connect:        0    0   1.5      0       7
Processing:   711 1847 520.7   1929    2755
Waiting:      699 1827 518.1   1898    2750
Total:        711 1847 520.7   1929    2755

Percentage of the requests served within a certain time (ms)
 50%   1901
 66%   2167
 75%   2216
 80%   2279
 90%   2525
 95%   2696
 98%   2736
 99%   2752
100%   2755 (longest request)

It is most likely that the combinations result in faster page rendering and that the user experince benefits from
all of these optimisations but I must admit ... I expected a different result and it seems that 1.4 version
needs significant higher server specs to do the same job.
Maybe I have overseen something and someone else gets different results but this post is not ment to complain about the new version than just sharing some interesting observations.

Update
So I tested the performance now with http://tools.pingdom.com to see how the user experience is and is quite obvious that
the 1.4.1 with optimisations performs better and results in faster page loading.... so my conclusion is.
1.4 is better for the user experience but you need more server power to compensate the higher cpu usage.

Best regards, trip

[Martin C]

I wonder what server specs you must have to let PS run smootly?
Does anyone has experience with it?

Martin

[Trip]

Hi Mcrielaard,
the tests I made were on a desktop system and it simulates 10 users at the same time with is very rough. After I did those stress test
I tested the real site performance with pindom tools and in average the page loading was half the time than the old version which is great.
For my project the number of request per page load reduced from about 100 to 30 request in the new version. This is really cool but needs (and due to other enhancements) more cpu power.
I run my sites on a root server and is not a big deal to serve the new version.
In europe I've seen virtual cloud servers with root access for about 9€ a month and there were some adds in the forum with optimized prestashop hosting that were not to expensive.
I think when you stick with 3 € homepage hoster which serves hundrets of other sites at the same time than it might be problemetic depending on the hoster but I am confident that you can find premium space for 10€ or less and if your hoster does not provide the power than the problem is with the hoster and not with prestashop.
Best regard, trip