Browse Prior Art Database

Efficient Handling of Queued Read/Write Requests

IP.com Disclosure Number: IPCOM000036579D
Original Publication Date: 1989-Oct-01
Included in the Prior Art Database: 2005-Jan-29
Document File: 2 page(s) / 13K

Publishing Venue

IBM

Related People

Williams, JS: AUTHOR

Abstract

A method is described for efficient handling of sequential disk requests issued to the Extended ESDI adapter care. This method provides for significantly greater performance than was obtained by the prior art implementation. The prior method did not address a significant function with respect to handling read/write requests that build up on the device driver input queue.

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

Page 1 of 2

Efficient Handling of Queued Read/Write Requests

A method is described for efficient handling of sequential disk requests issued to the Extended ESDI adapter care. This method provides for significantly greater performance than was obtained by the prior art implementation. The prior method did not address a significant function with respect to handling read/write requests that build up on the device driver input queue.

The prior method addresses a situation where the device driver processes disk requests faster than the operating system requests. This resulted in the device driver queue being empty a great percentage of the time. The aforementioned implementation addresses this phenomenon by allowing the device driver to read ahead into a cache on the ESDI adapter card. When the requests eventually come, they are serviced out of the cache.

A situation occurs where the queue can build up with many sequential requests. In the case of sequential read requests, the previous implementation processed them the same as if there were only a few requests. This resulted in a performance of roughly 500k bytes per second. In the case of write requests, the previous implementation did not address the situation and the resulting performance was less than previous file attachments.

Since the requests were already present on the queue, it is possible to service them at the native device speed of 900k bytes per second. The implementation of this design allows for performance for reads and writes of approximately 900k bytes per second.

This design is predicated on the capability of the adapter to accept two commands at a time. One to work on currently, and the next sequential request to work on when the current one is complete. This allows for the adapter to move to the next work request without missing a revolution of the disk.

The design is as follows:...