Planning for and enabling collocation

Understanding the effects of collocation can help reduce the number of media mounts, make better use of space on sequential volumes, and improve the efficiency of server operations.

About this task

Table 1 lists the four collocation options that you can specify on the DEFINE STGPOOL and UPDATE STGPOOL commands. The table also shows the effects of collocation on data that belongs to nodes that are and are not members of collocation groups.

Table 1. Collocation options and the effects on node data
Collocation option	If a node is not defined as a member of a collocation group	If a node is defined as a member of a collocation group
No	The data for the node is not collocated.	The data for the node is not collocated.
Group	The server stores the data for the node on as few volumes in the storage pool as possible.	The server stores the data for the node and for other nodes that belong to the same collocation group on as few volumes as possible.
Node	The server stores the data for the node on as few volumes as possible.	The server stores the data for the node on as few volumes as possible.
Filespace	The server stores the data for the node's file space on as few volumes as possible. If a node has multiple file spaces, the server stores the data for different file spaces on different volumes in the storage pool.	The server stores the data for the node's file space on as few volumes as possible. If a node has multiple file spaces, the server stores the data for different file spaces on different volumes in the storage pool.

Table 2. Collocation group options and effects on file space data
Collocation option	If a file space is not defined as a member of a collocation group	If a file space is defined as a member of a collocation group
No	The data for the file space is not collocated.	The data for the file space is not collocated.
Group	The server stores the data for the filespace on as few volumes in the storage pool as possible.	The server stores the data for the file space and other file spaces that belong to the same collocation group on as few volumes as possible.
Node	The server stores the data for the node on as few volumes as possible.	The server stores the data for the node on as few volumes as possible.
Filespace	The server stores the data for the node's file space on as few volumes as possible. If a node has multiple file spaces, the server stores the data for different file spaces on different volumes in the storage pool.	The server stores the data for the file spaces on as few volumes as possible. If a node has multiple file spaces, the server stores the data for different file spaces on different volumes in the storage pool.

When deciding whether and how to collocate data, complete the following steps:

Procedure

Familiarize yourself with the potential advantages and disadvantages of collocation, in general. For a summary of effects of collocation on operations, see Table 1.
If the decision is to collocate, determine how the data is to be organized, whether by client node, group of client nodes, or file space. If the decision is to collocate by group, you must decide how to group nodes:
- If the goal is to save space, you might want to group small nodes together to better use tapes.
- If the goal is potentially faster client restores, group nodes together so that they fill as many tapes as possible. Doing so increases the probability that individual node data will be distributed across two or more tapes and that more tapes can be mounted simultaneously during a multi-session No Query Restore operation.
- If the goal is to departmentalize data, then you can group nodes by department.
If collocation by group is the wanted result:
1. Define collocation groups with the DEFINE COLLOCGROUP command.
2. Add client nodes to the collocation groups with the DEFINE COLLOCMEMBER command.
The following query commands are available to help in collocating groups:
QUERY COLLOCGROUP

Displays the collocation groups defined on the server.

QUERY NODE

Displays the collocation group, if any, to which a node belongs.

QUERY NODEDATA

Displays information about the data for one or more nodes in a sequential-access storage pool.

QUERY STGPOOL

Displays information about the location of client data in a sequential-access storage pool and the amount of space a node occupies in a volume.

You can also use Tivoli® Storage Manager server scripts or PERL scripts to display information that can be useful in defining collocation groups.
Specify how data is to be collocated in a storage pool with the COLLOCATE parameter on the DEFINE STGPOOL or UPDATE STGPOOL command.
If you decide later that you want to delete members of a collocation group, you can use the DELETE COLLOCMEMBER command. You can also update the description of a collocation group with the UPDATE COLLOCGROUP command and delete entire collocation groups by issuing the DELETE COLLOCGROUP command.

Results

Tip: If you use collocation, but want to reduce the number of media mounts and use space on sequential volumes more efficiently, you can:

Define a storage pool hierarchy and policy to require that backed-up, archived, or space-managed files are stored initially in disk storage pools.
When files are migrated from a disk storage pool, the server attempts to migrate all files that belong to the client node or collocation group that is using the most disk space in the storage pool. This process works well with the collocation option because the server tries to place all of the files from a particular client on the same sequential-access storage volume.
Use scratch volumes for sequential-access storage pools to allow the server to select new volumes for collocation.
Specify the client option COLLOCATEBYFILESPEC to limit the number of tapes to which objects associated with one file specification are written. This collocation option makes collocation by the server more efficient; it does not override collocation by file space or collocation by node.

When creating collocation groups, keep in mind that the ultimate destination of the data that belongs to nodes in a collocation group depends on the policy domain to which nodes belong. For example, suppose that you create a collocation group that consists of nodes that belong to Policy Domain A. Policy Domain A specifies an active-data pool as the destination of active data only and has a backup copy group that specifies a primary storage pool, Primary1, as the destination for active and inactive data. Other nodes in the same collocation group belong to a domain, Policy Domain B, that does not specify an active-data pool, but that has a backup copy group that specifies Primary1 as the destination for active and inactive data. Primary1 has a designated copy storage pool. The collocation setting on PRIMARY1, the copy storage pool, and the active-data pool is GROUP.

When the node data is backed up and a simultaneous-write operation occurs, active and inactive data is stored in Primary1 and the copy storage pool. Note, however, that although all the nodes belong to a single collocation group, only the active data which belongs to nodes in Domain A are stored in the active-data pool. The data in Primary1 and the copy storage pool is collocated by group. The data in the active-data pool is also collocated by group, but the "group" consists only of nodes that are members of Policy Domain A.