Browse Prior Art Database

Improving Cache Performance in Swapping Environments

IP.com Disclosure Number: IPCOM000050164D
Original Publication Date: 1982-Sep-01
Included in the Prior Art Database: 2005-Feb-10
Document File: 2 page(s) / 13K

Publishing Venue

IBM

Related People

Operowsky, HL: AUTHOR

Abstract

The purpose of any high speed DASD (direct-access storage device) cache is to minimize delay in accessing frequently referenced data. It is therefore important to provide space in the cache for such data by eliminating unwanted items as soon as possible. Unfortunately, it is not always easy to determine which items are no longer needed, and even when it can be done, the cost may be extremely high.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 53% of the total text.

Page 1 of 2

Improving Cache Performance in Swapping Environments

The purpose of any high speed DASD (direct-access storage device) cache is to minimize delay in accessing frequently referenced data. It is therefore important to provide space in the cache for such data by eliminating unwanted items as soon as possible. Unfortunately, it is not always easy to determine which items are no longer needed, and even when it can be done, the cost may be extremely high.

MVS (multiple virtual system) "swaps" or transfers a user out of real memory when it deter mines that the resources consumed by that user can be used more profitably elsewhere. As system conditions change, swapped out users are swapped back into real memory and allowed to execute. The storage used by that user, his working set, is typically written to "swap data sets" on DASD which are preformatted into 96K-byte blocks called "swapsets". MVS allocates these swapsets to users when swapping them out and makes the swapsets available again after swapping the user back in. Once the swapin operation is complete, there is never a need to refer back to the swapset.

Swapping is most often employed by MVS in TSO (time sharing option) environments. After the system completes a user's requests, it swaps that user out and swaps him back in when he enters his next command. It is therefore important to complete the swapin as quickly as possible in order to minimize the user's response time. If a high speed cache were used to hold swapsets, swapin time could be substantially decreased. But to be effective, the cache must be ...