Browse Prior Art Database

Record Caching Scatter Index Table Directory Structure

IP.com Disclosure Number: IPCOM000120975D
Original Publication Date: 1991-Jul-01
Included in the Prior Art Database: 2005-Apr-02
Document File: 6 page(s) / 236K

Publishing Venue

IBM

Related People

Legvold, VJ: AUTHOR [+3]

Abstract

A 3990 Control Unit that provides the Record Caching function also requires a subset of the Track Caching function as provided by the standard 3990 Model 3.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 33% of the total text.

Record Caching Scatter Index Table Directory Structure

      A 3990 Control Unit that provides the Record Caching
function also requires a subset of the Track Caching function as
provided by the standard 3990 Model 3.

      Record Caching was previously provided with the 3880. The
directory in this control unit is based upon a Scatter Index Table
(SIT) which points to individual Record Directory Entries (RDE).  The
Record Directory Entry contains information identifying the record
and its status, pointers to the record image in cache, forward and
backward pointers for its MRU/LRU Active List and a hash chain
pointer (see Fig. 1).

      The 3990 Track Caching directory is also based on a SIT
structure, but points to Track Directory Entries (TDE), which contain
information pertaining to all records on the same track.  The TDE
also identifies the track and its status, has forward and backward
MRU/LRU Active List Pointers, and a hash chain pointer (see Fig. 2).

      The pointers to the track image in cache, however, are located
in a Track Information Block (TIB), which contains a summary of where
all the individual records are located in cache and/or Non-Volatile
Storage (NVS).

      A Track Slot Header (TSH), which is paired with a unique TDE,
contains information as to which and how many segments are used for
the track image, and which of these segments contain the variable
sized TIB.  The TDE/TSH pair is made up of two consecutive cache
words, each with its own Error Correction Bytes (SRC) so that they
can be accessed independently by different operations.
PROBLEM

      The 3990 Record Caching control unit requires both the record
caching and track caching directory information.

      The problem is to perform both Record and Track Caching
functions while minimizing the performance impact of searching for
data in two different formats.
PROPOSED SOLUTION

      Instead of using the Scatter Index Table (SIT) to point to
separate modes of directory entries, or to an additional structure
which then distinguishes between Record and Track directory entries,
the SIT now contains the information identifying the track and its
status.  The hash chain consists of a "Base" SIT entry which points
to successive SIT expansion (SITX) entries used to represent
additional tracks that hash to the same spot.  The SIT still contains
the RDE anchors, as did 3880 Record Caching, and a pointer for a TDE,
as with 3990 Track Caching, but these structures do not have to be
accessed for ALL operations, since the SIT contains a summary of the
record and track status information including the only "BUSY"
indicator for the entire track (see Figs. 3A and 3B).

      A hit/miss can now be determined on ONE cache access instead of
TWO.

      A directory scan can determine if a track qualifies for the
specified operation in ONE cache access instead of TWO.

      The TSH and TIB structure remain unchanged from 3990 Track...