Planning storage groups for data sets

Ideally, you would have only one storage group containing all of your data sets, and the system would manage everything. Realistically, you have to account for a variety of data sets, such as databases, large data sets, temporary data sets, and tape data sets.

Before you actually define your storage groups, you should gather the following information:

Existing I/O hardware configuration
Projected hardware requirements
Estimated general requirements of user groups
Anticipated security requirements
Backup and recovery requirements
Data set sizes
Required system access, based on both shared data requirements and the needs of user groups
The number of systems to which the storage group is connected

To begin your planning, define one primary storage group and identify all the data sets that do not fit into this general storage group category. Keep in mind your storage class and management class requirements.

For data sets that require continuous access (continuous for the storage class availability attribute) the storage group must contain sufficient volumes that do not interrupt data availability for a single device failure (for example, dual copy or array DASD).

The highest performance is generally obtainable through use of cache at the subsystem level, the device level, or both. Sufficient volumes with cache access should be available in a storage group for data sets that require high performance. Data sets that have primarily write access (write BIAS in the storage class) benefit most from a DASD fast write capability.

For data sets requiring concurrent copy, ensure that sufficient volumes in the storage group are attached through 3990 Storage Controls with Extended Platform, or through an IBM® RAMAC Virtual Array.

A stripe is the portion of a striped data set that resides on one volume. For striped data sets, ensure that there are a sufficient number of separate paths to DASD volumes in the storage group to allow each stripe to be accessible through a different path. The maximum number of stripes for physical sequential (PS) data sets is 59. For VSAM data sets, the maximum number of stripes is 16. Only sequential or VSAM data sets can be striped.

Figure 1 shows the formula that is used to derive the optimum number of stripes that are allocated to the data set:

Figure 1. Formula for Deriving Optimum Number of Stripes

Sustained data rate / Device transfer rate = Optimum number of stripes

where:

Sustained data rate is the target throughput rate, specified in the storage class definition
Device transfer rate is based on the published characteristics of the device

Ideally, the number of stripes should be on different paths, however, if this is not possible, SMS allocates to a smaller number of stripes.

If a data set has a management class that specifies automatic backup or migration, you must direct the data set to a storage group that is eligible to be processed for automatic backup or migration. You need to specify through ISMF the system or system group that is to perform the backup or migration only if that particular system or system group should do the processing. Otherwise, do not specify a particular system or system group to do the processing. It is done automatically. Management class attributes do not apply to tape data sets, so assign data sets to tape only if they do not require management class services.