Re: Database Clustering on multiple harddisk

["Alex Turner" <armtuk(at)gmail(dot)com>]

On 6/20/06, arie nugraha <dicarve(at)gmail(dot)com> wrote:

Thanks a Alex,

But can tou give a bit explanation about multiple separate tablespace

This will give you the basic syntax for the commands.
http://www.postgresql.org/docs/8.1/static/sql-createtablespace.html
http://www.postgresql.org/docs/8.1/static/sql-createtable.html

A tablespace is a designated location on a parition to store database objects such as tables and indexes.  They are often used to spread a large databases over multiple logical partitions to improve performance.

and HBA,

The host bus adapter is what connects your computer to your storage/discs. Often a SCSI, iSCSI, SAS. Fibre Channel or a SATA card depending on your storage type.  This card controls how your disc is configured and set up with regard to RAID levels (If you have a RAID capable card) (Redundant Array of Independant Discs, typically mirroring/striping or striping with parity RAID 1, RAID 10 or RAID 5)
http://en.wikipedia.org/wiki/RAID talks more about RAID levels and what they mean.

RAID arrays are typicaly transparent to your OS, so your OS doesn't care about the RAID level, only the HBA does.

RAID allows you to configure redunancy in your disc array, so that if one disc fails, there are others with the data on it so that your server doesn't go down and you don't loose data (which is what I suspect you want).  A RAID array is managed purely by the HBA (at least in the case of hardware RAID, which is what you want because it supports Battery Backed Caches), and doesn't affect how the OS sees the disc.  And seeing that postgresql makes calls to disc through the OS, if the OS doesn't care, neither does postgresql.  The thing you have to be carefull of is drivers for your HBA in the OS.  Many HBAs have very poor drivers, especially for Linux and other UNIXes, so you can often see very bad performance because of a poor HBA driver.

or is there any resources, tutorial on how to use this features on
postgresql

Thanks a lot


Alex Turner wrote:
> I think you may be a little confused...
>
> Last time I checked, Oracle RAC doesn't actualy support clustering the
> I/O, it's one Database backend with multiple instances on the
> front-end, which is only clustering the CPU bound part (it's also VERY
> expensive).
>
> If by clustering you mean multiple seperate tablespaces, then
> Postgresql does support them.
>
> If you mean RAID, then that is implemented by the HBA (host bus
> adapter), and is transparent to Postgresql.
>
> If you are looking for replication, many people find Slony a good add-on.
>
> Alex
>
> On 6/20/06, *Arie Nugraha* <dicarve(at)gmail(dot)com
> <mailto:dicarve(at)gmail(dot)com>> wrote:
>
>     Hi list,
>
>     I am new postgresql user and i had read the documentation but
>     i dont if PostgreSQL support database clustering on multiple harddisk
>     (multiple I/O, just like DB2 and Oracle does)
>     or not.
>
>     If postgreSQL support the database clustering is there any
>     documentation or tutorial that explain how to do it?
>
>     Thanks all.
>
>


---------------------------(end of broadcast)---------------------------
TIP 3: Have you checked our extensive FAQ?

               http://www.postgresql.org/docs/faq