Kohana vs Codeigniter, Performance Synthetics

I want to give a comparison of the performance of two amazing (and, in my humble opinion, delightful) frameworks, Kohana and Codeigniter . The test is synthetic, but, it seems to me, still interesting.

Test conditions

In both frameworks, I did not use the view, but only the work inside the controllers. DB: Postgresql 8.3 on localhost. Both controllers have the simplest form.
In the first test, there was a simple loop in 333 iterations, inside of which a call was made to the database on the following request:

SELECT *
FROM tbl_clients
JOIN tbl_clients_recvisites ON tbl_clients.clientid = tbl_clients_recvisites.clientid
JOIN tbl_recvisites ON tbl_clients_recvisites.recvisiteid = tbl_recvisites.recvisiteid
JOIN tbl_clients_contacts ON tbl_clients.clientid = tbl_clients_contacts.clientid
JOIN tbl_contacts ON tbl_clients_contacts.contactid = tbl_contacts.contactid

* This source code was highlighted with Source Code Highlighter .

Additionally, a profiler built into both controllers was included.
In the second test Apache JMeter was used.

Test One - Pure Synthetics Without Emulation

On each of the frameworks, I repeated the actions 10 times (10 times [F5]) and recorded the results of the application execution (delivery time is not taken into account - only PHP works). The results in the table:

	Codeigniter	Kohana
one	26,6959	14,808
2	26.6049	14,178
3	27,5487	14,289
four	26,7427	13,845
five	26.6099	14,554
6	26,5644	14,381
7	26,9622	13,717
eight	26,8828	14,242
9	27.2445	13,771
ten	26,8514	14,479
The average	26.87	14.23

And further. Important note : Codeigniter consumed 1.74 MB of RAM in this test, and Kohana - 1.35 MB

Test Two - Approaching Reality

This test used JMeter to simulate simultaneous requests.
The test consisted in the following. Each of the subjects (CI and KOH) sent 10 simultaneous streams of 5 requests each (in the stream they will be performed in turn), but not to “kill” the stand, the number of queries to the database was reduced from 333 to 9 (well Yes, I love the number 3;))
In the first case, the frameworks gave the profiler data, but they turned out to be different in the amount of bytes transferred, so I conducted the second series of tests when both frameworks gave out only 2 bytes ('ok') after the work was done. Each test was run in JMeter 3 times.
Summary Tables:
Codeigniter 9 requests profiler enabled

Average (ms)	Min (ms)	Max. (ms)	Throughput (queries / sec)	Kb / s
3880	2492	4370	2.5	57.17
3729	2538	4105	2.6	59.5
3760	2608	4263	2.6	59.04

Kohana 9 requests profiler enabled

Average (ms)	Min (ms)	Max. (ms)	Throughput (queries / sec)	Kb / s
2495	1104	2990	3.9	27.62
2462	1239	2873	3.9	27.89
2485	933	2955	3.9	27.68

Codeigniter 9 requests, output 'ok'

Average (ms)	Min (ms)	Max. (ms)	Throughput (queries / sec)	Kb / s
3749	2778	4241	2.6	0.01
3691	2187	4111	2.6	0.01
3776	3041	4208	2.6	0.01

Kohana 9 requests, output 'ok'

Average (ms)	Min (ms)	Max. (ms)	Throughput (queries / sec)	Kb / s
2451	1035	2870	3.9	0.01
2445	997	2924	3.9	0.01
2448	1266	2883	3.9	0.01

As can be seen from these tables, the number of bytes returned during the operation of applications did not have any effect.
')

findings

Everyone makes conclusions for himself, I just decided to test such similar, but at the same time, different frameworks. For myself, I think I did it.
Never forget that for every job there is the most suitable tool for it ...