Browse Prior Art Database

Empirical Method of Selecting Objects to Stage Into a Cache

IP.com Disclosure Number: IPCOM000036681D
Original Publication Date: 1989-Oct-01
Included in the Prior Art Database: 2005-Jan-29
Document File: 2 page(s) / 14K

Publishing Venue

IBM

Related People

Cochrane, PC: AUTHOR

Abstract

Disclosed is a mechanism which compares the performance of providing objects from their home locations with providing them from a cache, and recommends which objects should be staged into the cache. The mechanism continuously gathers samples of performance statistics and calculates relative staging values for the objects in each sample. The staging value of an object is the sum of four independent factors which are weighted according to their importance. This mechanism dynamically adapts to changing conditions, reduces the amount of storage needed for the cache, reduces the costs of copying objects into the cache, and minimizes dependencies on the implementations of the cache and the objects' home location(s).

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

Page 1 of 2

Empirical Method of Selecting Objects to Stage Into a Cache

Disclosed is a mechanism which compares the performance of providing objects from their home locations with providing them from a cache, and recommends which objects should be staged into the cache. The mechanism continuously gathers samples of performance statistics and calculates relative staging values for the objects in each sample. The staging value of an object is the sum of four independent factors which are weighted according to their importance. This mechanism dynamically adapts to changing conditions, reduces the amount of storage needed for the cache, reduces the costs of copying objects into the cache, and minimizes dependencies on the implementations of the cache and the objects' home location(s).

The mechanism has two asynchronous processes: a statistics-gathering process and a staging-evaluation process. First, the staging- evaluation process initializes a sample collection area and marks it active. For each object provided to a user, the statistics-gathering process records the fastest and slowest times to provide the object from home and from the cache and the number of times the object is provided from home and from the cache. When a significant number of statistics have been collected (2000 objects provided or one object provided 10 times from home) or when directed by an optional installation exit, the statistics- gathering process notifies the staging-evaluation process. The staging- evaluation process switches the statistics-gathering process to use an alternate sample collection area and then analyzes the just completed sample. The staging-evaluation process uses values derived from statistics for similar objects to fill in any absent statistics.

For each object which is already staged or which has been recently used, the staging-evaluation process calculates three independent factors of the savings that would be obtained if the object were staged: response time savings, contention savings and storage savings. These three factors are defined below. The staging-evaluation process normalizes each savings factor into a relative number in the range - 10,000 to +10,000 by multiply...