Browse Prior Art Database

Bandwidth based load balancing in EtherChannel Disclosure Number: IPCOM000236590D
Publication Date: 2014-May-05
Document File: 3 page(s) / 38K

Publishing Venue

The Prior Art Database


A method for bandwidth based load balancing in EtherChannel is disclosed.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 3

Bandwidth based load balancing in EtherChannel

Disclosed is a method for bandwidth based load balancing in EtherChannel.

EtherChannel or Link aggregation is a networking technology which allows several physical network adapters to be bundled together and distribute network traffic based on a hash algorithm. This provides aggregated bandwidth and redundancy in case of physical adapter failure. The current EtherChannel framework allows physical network adapters of different capacity to be bundled together, but fails to consider the capacity of the underlying physical network adapter when distributing traffic. For example, it allows 1 GigE and 10 GigE adapters to bundle together but then treat them equally when distributing network traffic. As a result, the higher capacity adapters e.g. 10 GigE are under utilized compared to lower capacity (e.g. 1 GigE) adapters. This results in the physical network adapters to be under utilized and also workloads using EtherChannel would see lesser throughputs than the aggregated bandwidth. Disclosed is a solution to address these problems such that the capacity of underlying physical network adapter is considered when distributing the traffic which helps to improve physical network adapter utilization as well as getting higher throughput for workloads.

The disclosed method utilizes a load balancing policy which assigns relative weights to each physical network adapter in EtherChannel based on its capacity. Traffic is then distributed based on the relative weight of the physical adapter when packets are transmitted over EtherChannel. Higher relative weight means the adapter will receive more traffic compared to lower relative weight adapters. Such a traffic distribution can be achieved by combining the weight of the physical adapter with the existing hash algorithm.

For example, if EtherChannel is composed of a 1GigE adapter and a 10GigE adapter, 1G adapter will have a weight of 1 and 10G will have a weight of 10. Then when a traffic is distributed based on the hash algorithm, the 10GigE adapter will receive traffic 10 times more than 1GigE adapter. For round robin hash algorithm, this means for every 10 packets sent over 10G, 1 packet will be sent over 1G. Similarly for src_dst_port hash algorithm, for 10 connections created over 10G adapter, 1 new connection will be created over 1G adapter. This way the traffic distributed to 10G adapter is 10 times more than the 1G adapter and better adapter utilization is achieved.

Advantages o...