[TYPO3] 2 Questions on MySQL for Typo3

[Ries van Twisk]

On May 10, 2007, at 4:54 AM, Corrado wrote:

> Georg Großberger wrote:
>
>> Hi all,
>>
>> Yesterday I read Dmitrys presentation sheet on Typo3 perfomance  
>> tuning. He
>> suggests a RAID 5 for MySQL, but MySQL themselves recommend a RAID  
>> 10.
>> What are your experiences?
>>
>> Another thing is the partitioning feature in MySQL 5.1, which highly
>> improves performance, when working with a huge mass of records.  
>> How could
>> someone use this feature in Typo3? Partitioning via language, or  
>> PID or
>> something else?
>>
>> I'm everything but a MySQL guru, so someone may share his/her  
>> experiences
>> with us.
>>
>> Georg
>
> Concerning hardware, it depends on your budget. RAID 5 is a good  
> compromise
> between costs and safety of data. Mirroring is more expensive but  
> safer,
> and should also be faster, if it is done at hardware level. Be careful
> between software and hardware raid options, because hardware raid  
> option
> are effectively faster, whilst software raid option can put load o  
> your
> processing capability.
>
> That is what I can remember by now about raid.
>
> Concerning MySQL itself, why don't you drop a line on the mysql  
> list? they
> should know!
>
> Hope this helps
>
> Corrado
> -- 
> eeos communications (technical support)
> http://www.eeos.biz
> _______________________________________________
>


hey Corraddo (eeos on #typo3?),
I am not much of a MySQL guru and I know more about other databases  
and it's tuning.

However a lot of databases allow you to create table spaces, you  
should check if
MySQL also supports tablespaces. Using table spaces you can create  
different
locations for your databases and indices.

This is my setup on a nonmysql database.

I have two RAID 1 sets 2 controllers and 4 HD's (each have a hot spare).

One one RAID1 set I store my root FS and my dataset, and on the other
I store my indices.

So my indices and dataset is physically separated. Then the RDMB can  
do it's
lookup's on the indices and fetch data from the other array and send  
it to the application.

To give you an idea of speed.

A join of 5 tables where the largest table is 450.000 records, 3  
tables are 50.000 records
and a couple of small tables of 500 records, where the result is a  
couple of hundred records
takes about 250ms to return data directly after the RDBM is started.  
Additional requests to
the same data set get's done in around 5ms. Additional requests are  
non-cached from teh RDBM's
perspective and only because data is already found in disk caches.

So if you really have the need for speed... check if you can separate  
your indices from your data storage
on MySQL. Dmitry is a MySQL guru might now how to do that.


Ries