Apparatus and method for priority queuing with segmented buffers

Abstract

Apparatus and methods for efficient queuing and dequeuing using segmented output buffers comprising sub-buffers and priority queues. Output buffers are monitored for empty sub-buffers. When a newly empty sub-buffer is discovered, a refill request is enqueued in a ranked priority queue wherein the rank of the destination priority queue is based on the number of empty-sub-buffers in the requesting output buffer. All high priority refill requests are dequeued before lower priority refill requests, thereby reducing the possibility of starvation. Optionally, by using simple dequeuing criteria, such as a FIFO discipline, instead of complex algorithms designed to improve fairness, system resources may be conserved thereby improving system throughput.

Description

RELATED CASES[0001]This application is a continuation of U.S. pending application ser. No. 10 / 927,801 filed Aug. 26, 2004, the entire disclosure of which is incorporated herein by reference.FIELD OF INVENTION [0002]Invention relates to apparatus and methods for reducing the risk of underflow in a multiple queue system.BACKGROUND OF INVENTION [0003]Queues are used in a variety of applications, especially where resources are shared. For example, queues may be used to manage traffic in routers, file and data servers or to schedule processes and / or threads in single and / or multi-processor systems. A variety of techniques may be used to prioritize resource requests in an effort to improve throughput or efficiency or to guarantee a quality of service (QoS). However, many of these techniques run the risk of starving some requesters. In order to address the issue of starvation, complex algorithms may be used to enqueue and / or de-queue resource requests based on a variety of parameters. Howe...

AI Technical Summary

Benefits of technology

[0005]The present invention provides apparatus and methods for efficient queuing and dequeuing using segmented output buffers and priority queues. According to the current invention, each output buffer is segmented into a multiplicity of sub-buffers. A set of ranked priority queues are used to hold refill requests for the output buffers. Output buffers are monitored for empty sub-buffers. When a newly empty sub-buffer is discovered, a refill request is enqueued in a priority queue wherein the rank of the destination priority queue is based on the number of empty-sub-buffers in the requesting output buffer. By dequeuing refill requests in an order according to the rank of the priority queues, all high priority refill requests are dequeued before lower priority refill requests. In this way, the possibility of starvation is reduced because an output buffer's refill request will eventually be registered in the highest priority queue as it nears starvation.

[0006]According to an example of the current invention, a first-in first-out (FIFO) queuing discipline may be used to dequeue each priority queue. By using simple dequeuing criteria, such as a FIFO discipline instead of complex algorithms designed to improve fairness, system resources may be conserved thereby improving system throughput.

[0007]According to an example of the current invention, all sub-buffers may be the same size. Adding this constraint may improve the fairness of the current invention. In some cases, the items in the sub-buffers may be pointers or references and the size of the reference or pointer may not be related to the magnitude of the referenced items; in some cases, an additional constraint may be added wherein each referenced item has the same maximum magnitude.

Problems solved by technology

However, many of these techniques run the risk of starving some requesters.

However, when complex algorithms are used to reduce starvation, precious system resources are consumed by the increased calculations and / or delays are introduced, thereby reducing overall throughput.

Images