DASD Best Practices

See below for Mainframe DASD Best Practices developed by Larry Strauss in the Capacity Planning and Management team. By following these DASD practices we can help reduce our IPF DASD cost expense which has doubled since implementing XRC.

DASD Best Practices:
• Use system determined block size when writing sequential datasets
o Code BLKSIZE=0 in the JCL, the system will determine the optimal block size for the device being written to.
o COBOL programs should be coded BLOCK CONTAINS 0 RECORDS
o Optimal blocking will reduce the DASD space needed, save on I/O, run time, and CPU required to process data
o Free unused DASD when dataset is closed
o Code RLSE parameter on dataset space allocation to recover unused space
o Do not code RLSE on PDS datasets when member maintenance will require the allocated space
o VSAM datasets should be allocated using DATACLAS=EXT4GB
o Allows for extended format
o Allows for extended addressability
o LARGE DATASETS - To minimize allocation failure when allocating large datasets, and reduce the effect of DASD space issues. Allocate DASD extents no greater than 350 Cylinders. There is a higher likelihood of finding a smaller primary quantity on a volume than finding a volume that has say 1000 cylinders of free space.
o For example a job needs 1000 cylinders,
 DO NOT allocate SPACE=(CYL,(1000,100),RLSE)
 DO allocate SPACE=(CYL,(350,330),RLSE)
 SMALLER DATASETS - In general when requesting space for DASD datasets, allocate DASD extents no greater than 350 Cylinders. Allocate the space expected for the dataset as the primary quantity, and allocate the secondary quantity as about 10% of the primary.
o For example a job needs 300 cylinders,
 DO NOT allocate SPACE=(CYL,(30,30),RLSE)
 DO NOT allocate SPACE=(CYL,(300,300),RLSE)
 DO allocate SPACE=(CYL,(300,30),RLSE)
 Code optional BUFNO parameters for sequential and BUFND BUFNI parameters for VSAM
o Default buffers (5 for sequential datasets), (2 data and 1 index for VSAM datasets) are often inadequate. Use additional buffers especially when large numbers of records will be processed
o Additional buffers can reduce physical I/O, reduce run time and save CPU required to process the data
o Use SMB (system managed buffering) or BLSR (batch local shared resources) when appropriate on VSAM batch with high I/O

If you have any questions on these best practices you can contact Larry Strauss at 224-405-3059.

Thanks,
Ken Kieliszewski - AD Core Operations
Discover Financial Services
Work: 224-405-2510