Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Remote device awared aggregation group

IP.com Disclosure Number: IPCOM000235556D
Publication Date: 2014-Mar-10
Document File: 6 page(s) / 279K

Publishing Venue

The IP.com Prior Art Database

Abstract

This invention is trying to provide a method to protect the stack link bandwidth by hashing/distributing the traffic over the trunk on partner switch. A static/dynamic trunk on partner switch will be split into several sub trunks logically. Member ports connected to same unit of the stack switch will form one unique sub trunk. MAC addresses will be learned on sub trunks according to the ingress port of the traffic from stack switch instead of normal trunk. Through this way, known unicast traffic on partner switch will be hashed and forwarded over sub trunk instead of whole trunk.

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

Page 01 of 6

Remote device awared aggregation group

Distributed Multi-Link Trunk is distributed trunk with links on different stack members, Distributed Multi-Link Trunk Local Preference is a new hashing scheme that applies only to the distributed trunks in Stacking mode and its purpose is to protect the stack link bandwidth by hashing/distributing the traffic over the trunk, only on the those trunk's links that are physically present on the packet's ingress unit. If the packet's ingress unit does not have links in the trunk the hashing will be done normally respecting the rest of the hashing rules that are active.

Also this new hashing scheme is applied only to known unicast traffic in stack, the DLF (destination lookup failed), broadcast, multicast traffic will still be forwarded as before, respecting the initial hashing algorithm.

The draw back for such a hashing method is that thetraffic over a trunk may flow in both directions and Distributed Multi-Link Trunk Local preference can control only one direction (the egress from trunk in stack) while the trunk ingress is controlled by the partner switch so the incoming traffic over the trunk may still go over the stacking links. Then the stack link still may be oversubscribed also, it is illustrated in Figure-1.

Figure-1 Distributed Multi-Link Trunk local preference hashing for unicast traffic

Currently, there is no solution to solve such draw back.

This invention is trying to provide a method to protect the stack link bandwidth by hashing/distributing the traffic over the trunk on partner switch. A static/dynamic trunk on

1


Page 02 of 6

partner switch will be split into several sub trunks logically. Member ports connected to same unit ofthe stack switch will form one unique sub trunk. MAC addresses will be learned on sub trunks according to the ingress port of the traffic from stack switch instead of normal trunk. Through this way, known unicast traffic on partner switch will be hashed and forwarded over sub trunk instead of whole trunk, just as illustrated in Figure-2.

Figure-2 Sub Trunk is applied

We use a new trunk group table, FDB table, and introduce a new way for MAC learning, trunk building.

The new trunk group table is as below (Each trunk will be an entry of the trunk group table):
One entry of the Trunk Group

Port Num Port 0 Port 1 Port 2 Port 3

Sub Trunk 0

2



Page 03 of 6

Port 4

…...

Port 5 Port 6 Port...