filling of data volumes

[Former Member]

Hi all,

I have a question regarding the way maxdb allocates space in data volumes.

Is there a reorganization of the db data in background or does maxdb just "fill up" the available volumes.

For example:

1)

I create a database with 10 volumes and the space usage is increasing.

Does the volumes get filled up one after another (first fill volume 0001, then volume 0002, ...) or is there used a balancing over all available volumes?

2)

What does happen, if I now have to extend db space and add two new data volume?

Does maxdb now use only the new volumes or is there a reorganization which leads to an (almost) equal distribution of IO over all datafiles.

If there is no reorganization of db data, the only way to extend db space without loss in performance would be to add several data volumes or to do an backup/restore procedure?

Best regards,

Sascha

[henrik_hempelmann · Accepted solution]

> Hi all,

Hi Sascha,

> 1)

> I create a database with 10 volumes and the space

> usage is increasing.

> Does the volumes get filled up one after another

> (first fill volume 0001, then volume 0002, ...) or is

> there used a balancing over all available volumes?

the write load of the database is distributed over all attached data volumes to increase the write speed (it's best to have each volume on it's own i/o channel). So your data volumes will fill to the same level.

> What does happen, if I now have to extend db space

> and add two new data volume?

> Does maxdb now use only the new volumes or is there a

> reorganization which leads to an (almost) equal

> distribution of IO over all datafiles.

There is currently no redistribution of the data in place. The database still uses all available data volumes for writing, but the new (empty) volumes are preferred.

We are aware of the increased i/o load on the new data volumes, but the automatic balancing of data volumes is not available yet.

> If there is no reorganization of db data, the only

> way to extend db space without loss in performance

> would be to add several data volumes or to do an

> backup/restore procedure?

Backup/restore is one viable solution, but you can manually distribute the data by adding a lot more data volumes and deleting the old ones running in online mode as well.

regards,

Henrik

[Former Member · Accepted solution]

Hi Henrik,

so I could do an reorg with the following procedure?

1) add several new san disks

2) add data volumes with the space of the existing database + the space I want to extend the db

3) delete all "old" data volumes (on weekend?!?)

Do you know if there is a heavy influence on performance while deletion of a data volume?

regards,

Sascha

[henrik_hempelmann · Accepted solution]

Sascha,

the sequence is OK!

You will notice heavy I/O on the system during deletion of the volumes. It's will cost you a bit more performance than a standard 'check data'....

regards,

Henrik