Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Neural Network Policy Determination for Optical Volume Switching

IP.com Disclosure Number: IPCOM000107566D
Original Publication Date: 1992-Mar-01
Included in the Prior Art Database: 2005-Mar-22
Document File: 3 page(s) / 135K

Publishing Venue

IBM

Related People

Bigus, JP: AUTHOR [+4]

Abstract

A method for determining when to remove an optical disk from the disk drive using a Neural Network is disclosed.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 46% of the total text.

Neural Network Policy Determination for Optical Volume Switching

       A method for determining when to remove an optical disk
from the disk drive using a Neural Network is disclosed.

      There are several variables which affect the decision of when
to remove an optical disk from a disk drive.  These include:
      1.  The type of job making the last request - batch or
          interactive,
      2.  The number of batch and interactive requests on the
          queue,
      3.  The customer's expressed preference for batch or
          interactive,
          on a per drive basis,
      4.  The number of files now open on the volume,
      5.  The last operation performed on the volume - close
          vs. other,
      6.  The number of requests waiting for service,
      7.  The rolling average frequency of use for the
          volume,
      8.  The number of volumes in the library, and
      9.  The number of drives in the library.

      The importance of these parameters is described below.
      1.  The type of job last using the volume - batch or
interactive - is important because an interactive request tends to be
more random in nature.  However, batch jobs, in general, are much
less random and, hence, there is a higher probability that the next
request will utilize the same volume.
      2.  The number of batch and interactive requests on the request
queue is an important measurement because it is a good prediction of
the forthcoming frequency of random requests.
      3.  The customer's expressed preference for batch or
interactive is a configuration parm expressed as a percentage on a
per-drive basis.  (The default is 50 percent batch).  The value is
used to "weight" the predictive value of 2.  This is an important
element in the measurement scheme, as it allows customers to
influence, but not dictate the volume switching performance decision.
(Hopefully, thrashing and deadlock caused by customer malfeasance
will thus be avoided.)
      4.  The number of files now open per volume is an important
measurement because, in general, the flow of operations is to open a
file, read or write some data, and finally close the file.  If
multiple files are currently open this indicates a high probability
that additional read or write requests will be coming.
      5.  The last operation performed on the volume is important
because a close operation will be the last operation done for one
file.  For example, if the job is interactive and the file was just
closed, this greatly reduces the chances of another operation for
that same volume.
      6.  The number of requests waiting for service is important
because there is little or no need to switch volumes if no work is
waiting.  (In fact, when no work is waiting, after a nominal delay,
volumes are dismounted to save dismount tim...