[TYPO3-project-4-3] Testing the memcached backend

[georg kuehnberger]

Steffen Müller wrote:
> Hi.
> 
> On 15.03.2009 17:24 Ingo Renner wrote:
>> Steffen Kamper wrote:
>>
>>
>>> did anyone a measuring yet? I would be interested in performance
>>> compares.
>> you'll probably not notice too much of a speedup with a one-server
>> installation. Memcached is rather ment to speed up installations spread
>> over multiple machines
>>
> 
> I can confirm this.
> I already did some brief test with apachebench on a single server a
> while ago and didn't notice a difference between mamcached, filesystem
> and DB storage of cache_pages. The same results on a virtual machine:
> http://www.t3node.com/blog/testing-the-forthcoming-typo3-caching-framework-with-memcached/
> 
> Yesturday I did another test on the virtual machine with latest trunk
> and it was almost the same. (about 60 pages/s which a hello world
> content element.) Although I expected more throughput using memcached
> with less I/O bottlenecks on a virtual machine.
> I'll try to post a more detailed report next week.

Steffen, et al,
Given the nature of mysql-query-cache and memcached you wont experience 
any performance-boost as long as you only test a one page (as that one 
will be covered by mysql-query-cache in most cases.)

The benefits of memcached will pop-up as soon as you are benchmarking 
more pages than mysql-query-cache can hold in memory.

For any meaningful load-testing I recommend using jMeter which is able 
to do the above and even more.

regards georg