Browse Prior Art Database

Write Cache Improvement by Combination of Rotational Positioning Reordering and Skip Write

IP.com Disclosure Number: IPCOM000123912D
Original Publication Date: 1999-Jul-01
Included in the Prior Art Database: 2005-Apr-05
Document File: 2 page(s) / 53K

Publishing Venue

IBM

Related People

Nishimiya, H: AUTHOR [+5]

Abstract

Disclosed is a method to improve performance of cached write commands using not only Rotational Positioning Optimization (RPO) method but also Skip Write method.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 55% of the total text.

Write Cache Improvement by Combination of Rotational Positioning
Reordering and Skip Write

   Disclosed is a method to improve performance of cached
write commands using not only Rotational Positioning Optimization
(RPO) method but also Skip Write method.

   Hard Disk Drive reorders cached writes to improve
performance.  RPO is a method to find one command which needs the
shortest access time from current writing sector.  Skip Write is a
method to write two or more random write sectors continuously by
skipping not requested sectors.

   When a cached write sector B is very close to current
writing sector A, RPO method select other command than B.  It is
because overhead time to start B is greater than latency from A to B.
In other words, it takes more than 1 revolution time to start writing
B after writing A.  However, if Skip Write method is applied, both
sector A and B are written continuously without 1 revolution latency.
In addition, Skip Write method decreases number of seeks.  This means
total seek time is also decreased.

   In other words, if two or more cached writes are very
close to each other, Skip Write method is applied for those writes.
In other cases, RPO method is applied.

   Figure shows implementation example of cached write
reordering using disclosed method.
  Figure
  ------
  Cached Write Information data structure (CWI): CWI has elements
  as follows:
    NEXT : pointer to next cached write data information
           structure
    SLBA : start lba of the write
    ELBA : end lba of the write
  exe_wr  : exe_wr points currently executing CWI.
  scan_wr : scan_wr points currently checking CWI.
  best_wr : best_w...