Browse Prior Art Database

Limited 1/P-Track Search Algorithm for Optimal Blocksize

IP.com Disclosure Number: IPCOM000115772D
Original Publication Date: 1995-Jun-01
Included in the Prior Art Database: 2005-Mar-30
Document File: 4 page(s) / 159K

Publishing Venue

IBM

Related People

Alvarado, JT: AUTHOR [+4]

Abstract

Disclosed is a more efficient algorithm for the search of an optimal blocksize for logical records of fixed blocked records for a given record length and device type.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 36% of the total text.

Limited 1/P-Track Search Algorithm for Optimal Blocksize

      Disclosed is a more efficient algorithm for the search of an
optimal blocksize for logical records of fixed blocked records for a
given record length and device type.

The algorithm:
  o  maximizes the number of logical records stored on a track on
      DASD, resulting in improved DASD space utilization;
  o  uses a minimum of additional processor storage; and
  o  uses a minimum of additional CPU instruction cycles to derive
      the optimum blocksize.

      For the current models of D/T3380 and D/T3390 DASD, this
algorithm always produces the optimal blocksize.

      DFSMS/MVS supports the concept of blocking, that is, grouping
together a set of logical records of record format with fixed length
(RECFM = Fx) into a single physical record, called a block.  The
block is the physical record stored on DASD using one I/O operation.

      For RECFM = Fx, on each track, all logical records have the
same Logical RECord Length (LRECL), and all blocks have the same
length, called the blocksize (BLKSIZE).  Currently, SFSMS/MVS
supports
a maximum BLKSIZE of 32,760 bytes.

Blocking is used to improve:
  o  the space utilization of a track - Each block requires the
      overhead of a single prefix block descriptor field (address
      marker, count area, key area, gap and etc.) that describes the
      set of logical records in the block.  Unblocked records (i.e.,
      one block for each logical record) require the overhead of one
      block descriptor field for each logical record.
  o  the I/O performance when reading the track, since blocking
      reduces the number of I/O operations required to read in all
the
      logical records on a track.

      For RECFM = Fx, to maximize space utilization on a track and
I/O performance when reading the track, it is important that an
efficient BLKSIZE be selected, given a track size and LRECL.

When a user creates a data set, he may either:
  o  manually select a BLKSIZE; or
  o  request the system to determine a BLKSIZE.

User-Determined-Blocksize - The user many manually select a BLKSIZE
by performing arithmetic calculations or by using a table that
contains
recommended BLKSIZEs for LRECL values, given a specific track size.

      In addition, the user needs to be cognizant of the geometry of
the device, and any subsequent movement of the data set to a device
with a different track size may require recalculation of the BLKSIZE
and reblocking of the logical records.

System-Determined-Blocksize - With system-managed-storage support,
the user may request the system to select a blocksize, called the
System-Determined-Blocksize (SDB).

General Problem - Prior to DFSMS/MVS 1.1.0, the algorithm used by the
system to calculate the SDB, did not always generate the optimal
BLKSIZE, i.e., the blocksize resulted in storing the largest number
of logical records on the t...