Browse Prior Art Database

Select Interrupt Assertion of Egress Events

IP.com Disclosure Number: IPCOM000008440D
Publication Date: 2002-Jun-13
Document File: 2 page(s) / 75K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a method that enables a device driver to selectively request interrupts for transmit events based on egress loading. Benefits include reduced interrupts per packet ratio, reduced CPU utilization, and improved throughput.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 50% of the total text.

Select Interrupt Assertion of Egress Events

Disclosed is a method that enables a device driver to selectively request interrupts for transmit events based on egress loading. Benefits include reduced interrupts per packet ratio, reduced CPU utilization, and improved throughput.

Background

The throughput of high-speed I/O controllers is often limited by the available processor bandwidth. Processing efficiency can be measured by the Performance Efficiency (PE) ratio

(i.e. the throughput in Megabits/sec) divided by the percentage of CPU utilization required to process the throughput level. An industry goal is to average one CPU cycle per byte for I/O activity. To meet such an aggressive goal, task offloading, interrupt moderation schemes, and other innovations must be employed.   

Interrupts are expensive operations for a pipelined processor; however, egress operations must be completed in a timely manner. Interrupts are used in operating systems (such as Windows NT) where timer function granularities are not fine enough to complete high-speed I/O operations in a timely manner.

General Description

This disclosed method reduces the number of interrupts by selectively requesting interrupt generation for egress operations. The disclosed method bases the request for an interrupt on the current workload. In times of low workload, interrupts are used with every egress operation. This avoids the latencies of waiting for interrupts when egress request operations are occurring at a slow rate. Under high workload rates, the interrupts are requested less frequently (N is larger). This reduces the amount of overhead on the system processor when efficient operation is needed most. In this case, the latencies are not detrimental to performance because subsequent operations occur at a high frequently. 

Methods of workload determination are well known. These include a measure of backlogged events, CPU utilization, a...