Benchmarking results: Fapws3 rocks ;-)
william opensource4you
I've also compared Sqlite and Postgresql.
Context is:
----------------
- I'm using an Archlinux system with Linux 2.6.30 SMP Preempt
- HW is an AMD 64 dual core 4800 not overclocked (2487Mhz) with 2GB ram
- every run was made with the command "nice -n20 ab -n500 ...." (in my
blog I explain why using nice).
- I've used a complex Django website I've developped while ago. Sqlite
and Pgsql are loaded with the same set of data.
- Sqlite-3.6.15, Postgresql-8.3.7, Pound-2.4.4, Apache-2.2.11,
Mod_python-3.3.1, Django-1.0.2, Fapws3-0.3, Python-2.6.2,
ApacheBenchmark-2.3
- the results show the mean request per seconds (thus higher is better).
With Sqlite as backend, the results are:
--------------------------------------------------------------
c fapws3-0.3 1x Fapws3 + Pound 6x Fapws3 + pound
Apache + mod_python
10 13.76 14.77 23.61
10.08
30 14.32 14.64 23.89
9.88
100 14.81 14.75 24.70
NA
where c is the concurrency.
NA is because the machine crashed. I've tried it 3x with the same effect.
With PGsql as backend, the results are:
---------------------------------------------------------------
c fapws3-0.3 1x Fapws3 + Pound 6x Fapws3 +
pound Apache + mod_python
10 11.01 10.78 17.07
16.49
30 10.88 10.65 18.60
14.63
100 11.00 10.51 18.44
1.85
Conclusions:
--------------------
- Fapws3 rocks and stays stable
- Pound does not add latency (direct = 1x Fapws3 +pound).
- Pound has a good balancing mechanism and allow to improve the
overall performance
- With low number of concurrency Fapws3 and Apache are quite similare,
but when hits are coming we see problems with Apache.
- As expected Sqlite is a bit faster than Postgresql.
William
ps:
I've kept all ab results. Thus is someone would like to see other
results, I can provide them.
proteusguy
where is your blog?
-- Ben
On Jul 28, 6:59 pm, william opensource4you <william.o...@gmail.com>
wrote:
> I've also compared Sqlite and Postgresql.