Browse Prior Art Database

Method for Improving Utilization of a Small Cache

IP.com Disclosure Number: IPCOM000115684D
Original Publication Date: 1995-Jun-01
Included in the Prior Art Database: 2005-Mar-30
Document File: 2 page(s) / 32K

Publishing Venue

IBM

Related People

Miro, F: AUTHOR

Abstract

Disclosed is a method for improving the utilization of a relatively small software cache, such as a 48K cache, in a multi-thread operating system by giving the running thread of execution an additional time slice of execution after it has found data in the cache.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 100% of the total text.

Method for Improving Utilization of a Small Cache

      Disclosed is a method for improving the utilization of a
relatively small software cache, such as a 48K cache, in a
multi-thread operating system by giving the running thread of
execution an additional time slice of execution after it has found
data in the cache.

      Without this method, when the scheduled time slice of a thread
of execution ends, there is a high probability that, due to a high
level of cache contention, the thread of execution will lose the
cache, with its cached data.  The performance of the file system is
reduced as Input/Output (I/O) requests to the file system cannot be
resolved in the small cache.

      This method for improving cache utilization is achieved within
an interface between the file system and scheduler components.  The
file system sets a flag whenever the thread of execution gets data
through the cache.  The scheduler checks this thread-based flag and
resets it.  If the state of the flag is true when it is checked, and
if no higher-priority thread is ready, the current thread is
rescheduled to use the cache for an additional maximum time slice.
In this way, the running thread can satisfy a higher percentage of
I/O requests through the cache, before its cached data is flushed due
to the disk access of another thread.