Browse Prior Art Database

Optimize Log Structure File System Garbage Collection by Using Segment Aging

IP.com Disclosure Number: IPCOM000122953D
Original Publication Date: 1998-Jan-01
Included in the Prior Art Database: 2005-Apr-04
Document File: 2 page(s) / 87K

Publishing Venue

IBM

Related People

Cohn, O: AUTHOR [+4]

Abstract

A technique is disclosed for efficiently managing free space within Log Structure File System (LSFS). The technique selects the Segments to be emptied using a Segment based criterion rather than a record or track based criterion.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Optimize Log Structure File System Garbage Collection by Using Segment
Aging

      A technique is disclosed for efficiently managing free space
within Log Structure File System (LSFS).  The technique selects the
Segments to be emptied using a Segment based criterion rather than a
record or track based criterion.

      The usage of the proposed Segment based criterion ensures that
the skew in the distribution of live data across the log is large
enough over time and therefore the management of free space is
efficient. In  addition, a Segment based criterion imposes
significantly less space requirements upon the control data structure
needed to support the management of free space.

      Log Structure File System (LSFS) is a new approach for disk
storage management.  Its main idea is to serve all modifications to
disk (write operations) sequentially in a log-like structure.  An
essential part of any LSFS is a process that collects live data and
thereby frees  enough sequential space for new data.  This process is
called the Garbage  Collection process.

      A popular approach for the development of efficient Garbage
Collection processes is to use a Segmented LSFS.  In this approach,
the disk space is divided to equal size slices called Segments.  The
sequentiality of the log is preserved within a Segment but not across
Segments.  When relatively large Segments are used, the advantage of
having a better structure for the Garbage Collection process
justifies the minor degradation in write throughput.

      Garbage Collection processes for Segmented LSFS move live data
from several partially full Segments, to one empty Segment.  In order
to be efficient, the Garbage Collection process should ensure that
the skew in the distribution of live data across the log is large
enough over  time.  Thus, some criterion should be used in selecting
the candidate Segments for the Garbage Collection process.

      All the criterions (e.g., creation time, logical location) that
are used in existing Garbage Collection processes are based on a
record or track characteristics.  Thus, large control data structure
is required  to support their usage.

      The disclosed technique introduces a Segment based criterion
and a method for using it.  The proposed cr...