Browse Prior Art Database

Optimizing Inactive Optical Drive Activity

IP.com Disclosure Number: IPCOM000112945D
Original Publication Date: 1994-Jun-01
Included in the Prior Art Database: 2005-Mar-27
Document File: 4 page(s) / 127K

Publishing Venue

IBM

Related People

Goldfeder, ME: AUTHOR [+2]

Abstract

This article uses a scoring algorithm to select what activity should be done when an optical drive in an optical library finishes a read query. Scores are kept on an individual volume basis as a running total (moving window) and the highest score for a particular volume would determine what post-read-query activity would take place. Thus, an optical drive in a library may do one activity with one volume once the query is finished and a different activity with a different inactive volume. Inactivity is denoted by zero opened files in a mounted volume.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 48% of the total text.

Optimizing Inactive Optical Drive Activity

      This article uses a scoring algorithm to select what activity
should be done when an optical drive in an optical library finishes a
read query.  Scores are kept on an individual volume basis as a
running total (moving window) and the highest score for a particular
volume would determine what post-read-query activity would take
place.  Thus, an optical drive in a library may do one activity with
one volume once the query is finished and a different activity with a
different inactive volume.  Inactivity is denoted by zero opened
files in a mounted volume.

      The list of activities which may transpire, once an optical
drive goes inactive following the execution of the last query, are
shown below.  The selected activity can vary with the type of data on
each volume, the age of the data, and the particular workload
accessing the volume at a given time.

1.  Keep reading from the end of the queried file, anticipating a
    sequential read pattern.  A file "read-ahead" would be accessed
    much more quickly from cache rather than from the optical disk.
    This is called the ReadAhead algorithm.

2.  Seek to the physical middle of the optical disk, assuming that
    the next query will require a totally random seek.  This is
    called the Middle algorithm.

3.  Seek to the Center of Gravity of the last N addresses or tracks
    queried on the optical disk.  This is called the CenterGrav
    algorithm because the actuator would be positioned at the center
    of gravity of the active portion of the optical volume.  Thus,
    prepositioning the actuator can be done based on the likelihood
    of the next access rather than merely prepositioning to the
    physical middle of the volume.

      To maintain the running total of scores and running average of
addresses or tracks, an AxNxV matrix is needed.  This matrix could be
stored in the OLMSCB (Optical Library Manager System Control Block)
of the library control unit or in the device itself.

o   Parameter A is the number of competing algorithms.  In this
    article, A equals 3.  These are (1) ReadAhead, (2) preposition to
    physical Middle of volume, and (3) preposition to Center of
    Gravity of the addresses or tracks of last N accessed files.

o   Parameter N is the number of accesses in the running total or
    moving window of scores and the number of accesses in the running
    average of addresses or tracks.  It is preferred that N be a
    binary number, such as 4 or 8.  This way, the arithmetic mean of
    the last N addresses or tracks would be calculated in C under an
    OS/2 thread by summation and a bit shift to the left.  This
    binary division is much faster than long division.

o   V is the number of volumes in the library.  This is equal to
    twice the number of cartridges for dual sided media and optical
    drives with one...