Designing for DB2 Data

DB2® system and recovery data sets can benefit from dual copy service for the following reasons:

The directory and DB2 catalog are critical and cannot be duplexed by DB2.
The boot strap data set (BSDS) and the active logs can be duplexed by DB2. Fault-tolerant devices can provide additional protection from hardware outages.
If you currently place the BSDS and dual active logs on separate volumes, you can continue in SMS by setting the storage class Guaranteed Space attribute to YES.

Recommendation: Do not use the guaranteed space attribute to specify volumes to separate the partitions of a DB2 table space. You can avoid using the guaranteed space attribute in the following situations:

SMS volume selection algorithms normally spread data set allocations if you have a large number of volumes in a storage group.
You can use DFSMS striping for linear data sets.
With the IBM® RVA and the ESS, multiple logical volumes can be mapped to a physical volume due to their RAID architecture, volume capacity, and, if applicable, their log structured array architecture.
The IBM ESS has large cache structures and sophisticated caching algorithms. It is capable of providing a much larger throughput. Its capabilities of parallel access volume and multiple allegiance allow many concurrent accesses to the same data. Therefore, specific volume placement and data set separation used for performance reasons should no longer be required.

Table 1 shows the relationship between the data set, data type, and LLQ that identifies the data set in the storage class ACS routine.

Table 1. DB2 Data Sets Requiring High Availability
DB2 data set	Data set type	3rd-level qualifier
DB2 catalog	System	DSNDB06
DB2 directory	System	DSNDB01

The active logs can benefit from DASD fast write and cache services, because DB2 transactions can wait for logging before completing. The access reference pattern for the database can affect the caching or DASD fast write benefits for production databases. The DFSMS I/O statistics provide long term measurement of data set accesses, response components, and cache statistics. They can be used in application tuning or batch window reduction. You can benefit significantly from I/O priority scheduling in a mixed workload environment. For example, to achieve consistent response times for transaction processing, you can prioritize transaction processing reads above query reads and DB2 asynchronous writes.

Table 2 shows the relationship between the data set, data type, and LLQ that identifies the data set in the storage class ACS routine.

Table 2. DB2 Data Sets Having High Write Activity
DB2 data set	Data set type	Low-level qualifier
Active Log	Recovery	LOGX/Y
Boot Strap	System	BSDSX/Y

The access reference pattern for the database can affect the caching benefits for production DB2 databases. Table 3 summarizes this result.

Table 3. DB2 Data Sets Having High Read Activity
DB2 data set	Data set type	2nd-level qualifier
Production databases	Databases	DSNDBC/D

The sample configuration has the following storage class assignments:

Class name: DB2 data sets
DBCRIT: Directory, Catalog, BSDS
FASTWRIT: Active logs
FAST: DB2 databases

All other DB2 data sets are assigned the STANDARD storage class. Refer to Table 1 for the list of attributes associated with these storage classes.