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

Customization of a General Purpose Lock Manager for a Cluster File System

IP.com Disclosure Number: IPCOM000112825D
Original Publication Date: 1994-Jun-01
Included in the Prior Art Database: 2005-Mar-27
Document File: 2 page(s) / 50K

Publishing Venue

IBM

Related People

Devarakonda, M: AUTHOR [+2]

Abstract

Disclosed is a method of customization of a general purpose lock manager for an AIX*-based cluster file system. This file system is built by parallelizing AIX single-system file system, JFS, for cluster-coupled hardware. Customization is needed for two reasons: (1) for naming the file data objects; and (2) for preventing deadlocks in managing a fixed-size lock table due to dependencies among data objects.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 58% of the total text.

Customization of a General Purpose Lock Manager for a Cluster File
System

      Disclosed is a method of customization of a general purpose
lock manager for an AIX*-based cluster file system.  This file system
is built by parallelizing AIX single-system file system, JFS, for
cluster-coupled hardware.  Customization is needed for two reasons:
(1) for naming the file data objects; and (2) for preventing
deadlocks in managing a fixed-size lock table due to dependencies
among data objects.

      To uniquely identify each file data object, a four-tuple key is
used.  The four-tuple consists of a device number (which in turn
comprises of major and minor numbers), a node-id, an inode number,
and a page number.  The device number is unique within a node, and
represents a logical volume (which stores the file data in AIX).
Therefore, the device number and the node-id fields are used to
uniquely identify a logical volume within the cluster.  Further,
since an inode number uniquely identifies a file in a file system,
the first three parts of the tuple uniquely identify a file within
the cluster.  The page number field is used to identify a page within
a file, and this field is added to provide locking of files at the
page level.

      The lock manager is also customized to prevent deadlocks in
managing a fixed-sized lock table in the presence of dependencies
among files.  In a fixed-sized lock table, if the table is full and
if a new file needs to be added to the t...