Browse Prior Art Database

Loop detection in a Publish/Subscribe infrastructure Disclosure Number: IPCOM000239515D
Publication Date: 2014-Nov-13
Document File: 1 page(s) / 19K

Publishing Venue

The Prior Art Database


This invention will now send a special message back to the queue managers who sent the message so that they can alert the administrators at all the queue managers in the loop that they are in a loop

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 54% of the total text.

Page 01 of 1

Loop detection in a Publish/Subscribe infrastructure

A typical large Publish/Subscribe infrastructure will span the globe and consist of some form of groups, these groups will be joined together by some form of Pub/Sub link since the traffic, connectivity and administration of one massive group would be unmanageable in today's mobile world where such infrastructures are having to cope with millions of publishers and subscribers. As these infrastructures grew it was all too easy for administrators to create loops in the infrastructures so that a publish would end up being published back to the source of the publication and then be forwarded on again thus swamping the infrastructure as the publication would carry on being republished forever more. To alert administrators a system needs to be in place, such as a "publish fingerprint" which detects the looping publications. Unfortunately as these systems have grown the simple alerting at the point of detection is not sufficient.

    Expanding on the publish fingerprint solution, at the point of detection, rather than just sending an error message to the logs, we will also publish a "loop detected" message back to the queue manager who we have just received the publication from, to a unique loop error topic, so that each queue manager in the loop will send an error message to its error logs and forward on the "loop detected" message back through the chain of queue managers. Thus let's imagine in a current messaging system an administrator in Hong Kong creates a loop, potentially a queue manager in New York will be sending error messages to its logs, with the new scheme the administrator in Hong Kong will, shortly after creating a loop, be notified on the queue managers, which they have just linked together, that they have just created a loop.

    In exis...