Browse Prior Art Database

Group QoS for Applications

IP.com Disclosure Number: IPCOM000016364D
Original Publication Date: 2002-Nov-01
Included in the Prior Art Database: 2003-Jun-21
Document File: 1 page(s) / 36K

Publishing Venue

IBM

Abstract

The challenge that most business face is managing the available bandwidth that is shared between various groups of application based on specified policies. The policies specify the amount of bandwidth that a application group is allocated for its use. This could be a contractual agreement between ISPs or setup a system administrator. The challenge is then to manage the available bandwidth among these application groups such that no application group gets more bandwidth than what has been allocated.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 70% of the total text.

Page 1 of 1

Group QoS for Applications

The challenge that most business face is managing the available bandwidth that is shared between various groups of application based on specified policies. The policies specify the amount of bandwidth that a application group is allocated for its use. This could be a contractual agreement between ISPs or setup a system administrator. The challenge is then to manage the available bandwidth among these application groups such that no application group gets more bandwidth than what has been allocated.

Our idea aims to solve this problem at the application level. By controlling the rate at which the application (server) pumps data, we can ensure the Quality of Service and guarantee fairness among all the application groups.

The application group is associated with a policy which specifies the amount of bandwidth that is allocated to it. Each group of applications is assigned to a common token bucket for the entire group. Every new connection by any application in the application group has its file descriptor associated with common token bucket for the group. When the processes in the application group send data at a rate greater than the allocated rate, the tokens in the bucket gets depleted. Rather than queueing the data packets down in the stack (as in traditional approach) our approach targets the application itself. Writes on network file descriptors will proceed only if the associated token bucket is not empty. Therefore if the process...