Collocation is the process of grouping your nodes together (collocating) to help improve restore performance.

Over time as you add more and more nodes to your environment, the data for these nodes will slowly become more and more dispersed across your tapes as you perform backups.  When restoring the data if only a small number of files are required, this will not be a problem but when a restore is needed for a large number of files and those files have become dispersed across a large number of tapes then restores time will become an issue. I have personally experienced an issue when restoring a large volume that took over 2 days when not using collocation.

Instead of having every node’s data being placed on tapes at TSM’s discretion, collocation allows you to define groups of nodes to be stored together on your tapes.

As with everything there is a trade-off, collocation groups will require the use of more tapes but can significantly improve restore times.

Badly configured collocation will cause the consumption of very large amount of tapes.  And in my experience this has been the number one cause of the “Where have my all my scratch tapes gone” and “Why do I have so many filling tapes” , we will explore this later in the document.

Collocation is set at the Storage Pool so it is possible to have different levels of collocation based on Storage Pools.  As Collocation is really designed to help with Physical Tapes it only applies to Primary Sequential and Copy Storage Pools.

There are a number of difference types of collocation methods that can be used

NO – Collocation is disabled for the Storage Pool

FILESPACE – TSM Attempts to put each file space on its own tape

NODE – TSM Attempts to put each Node on its own tape

GROUP – TSM will put groups of Nodes on their own tapes as defined in the collocation groups.  If no collocation groups are defined then the storage pool reverts to Node Collocation.

NOTE:  If Group Collocation is set and there are no collocation groups defined or there are nodes not defined in any collocation group then you very quickly run out of scratch tapes and see a lot of filling tapes with not much data on them

Firstly let’s look at how to check to see if collocation is set.  From the Admin CLI issue the command

Q STGPOOL F=D and check the status of the collocate field

From TSM Studio open the Storage category and open the Storage Pools Dataview and the check the value in the Collocate column

To change the collocation method being used

Admin CLI:  UPDATE STGPOOL COLLOCATE=GROUP

From TSM Studio

Go to the Storage Pools dataview and right click on the storage pool(s) you want to change

and choose Collocation

Choose the collocation method and press Ok

The next thing to do will be create the collocation groups

From TSM Studio open the Configuration category and open the Collocation dataview

Right Click and Choose Add Group

Enter the Group Details and then press Ok

Next Add Members to the Group, From the Collocation Dataview right click the Group and Choose Edit Group Members

NOTE: A Node can only be a member of one collocation group

Simply move nodes in and out of the group and press Ok

Read Here for Trouble Shooting Tips

Recommendations:  If you have plenty of tapes and will never run out then use FILESPACE or NODE.

If Tapes are at a premium and you dont care how long a restore will take then dont use collocation

Otherwise you will need to find a happy medium between tape usage and restore times using the Group method.  I have seen a number of different methods for collocating node, such as grouping by application type or OS Type.  I suggest putting nodes with small amounts of data that does not change very often together and putting nodes with large amounts of data by themselves.  For the rest of the nodes try and keep each group from using more that 50 tapes for any node in the group ( this number will depend on the number of drives you have, how fast the library can load the tapes and how fast the drive can move to the required location on the tape) to ensure fast restore times.

I have seen collocation disabled for copypools as a lot of organizations already replicate the critical data to their Business Recovery sites using other methods

As with everything I say in these articles test, test end test again

NOTE:  Virtual Tape Libraries do not have the same limitations as Physical Tapes. As the mount times and seek times are negligible collocation can be disabled without affecting restore times.

Posted in: TSM Studio