Browse Prior Art Database

Flow Control Based on Delay in Queues Instead of Memory Constraint

IP.com Disclosure Number: IPCOM000122983D
Original Publication Date: 1998-Mar-01
Included in the Prior Art Database: 2005-Apr-04
Document File: 2 page(s) / 79K

Publishing Venue

IBM

Related People

Beaulieu, C: AUTHOR [+2]

Abstract

Disclosed is a system of Flow Control based on delay in queues of messages, and not on number of messages or bytes as found in most systems.

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

Flow Control Based on Delay in Queues Instead of Memory Constraint

      Disclosed is a system of Flow Control based on delay in queues
of messages, and not on number of messages or bytes as found in most
systems.

Generally, in network nodes some hardware adapters receive the flow
of data from the network, store the frames in a shared memory and
enqueue some information related  to the received frames in a set of
shared Queues (called Microcode Input Queues or MIQs), and the
so-called kernel software schedules the set of queues and when a
queue is not empty, dequeues the message and dispatches a certain
task (called application) to handle the message.

In such systems there is a need to have flow control mechanisms to
monitor the MIQs and to allow the applications to take specific
actions (like slowdown the input rate of messages).  Most mechanisms
currently in place monitor the queues of messages with a threshold of
number of messages queued (or even bytes queued) independently of the
delay, i.e. the time a message stays in the queue before being
processed.  The problem is that sometimes the number of messages
queued does not exceed the value that will trigger flow control to
take place whereas a not acceptable delay exists in the messages
treatment and is not detected.

It is proposed to monitor still the number of messages but also to
take into account the time a message is waiting in the queue in
average and takes appropriate action with the external world to slow
down traffic if the delay is above an acceptable value.  The proposed
method is applicable to any system handling tasks waiting in queues
before treatment and where actions need to be taken based on the
delay a task is waiting.

The idea is the following: the kernel will 'measure' the delay in a
MIQ by computing the elap...