GRANULAR TRAFFIC STEERING AND LOAD SHARING WITH LINK GROUPS (LAG) IN SEGMENT ROUTED NETWORKS
Publication Date: 2015-Aug-11
The IP.com Prior Art Database
Nagendra Kumar Nainar: AUTHOR [+3]
New Segment Routing Identifier (SID) types are defined for Link Aggregation group (LAG) members which are advertised in the Interior Gateway Protocol (IGP), thereby enabling the Ingress node to use them as/when needed. Also local protection is specified for LAG group members.
Page 01 of 4
GRAXXXXX TRAFFIC STEERING AND LOAD SHARING WITH LINK GROUPS (LAG) IN SEGMENT ROXXXX NETWORKS
Nagendra Kumar Nainar Rajiv Asati Xxxxxx M. Pignataro
CISCO SYSXXXX, INC.
New Segment Rouxing Identifier (SXX) types are defined xor Link Aggregation group (LAG) xembers which arx xdvertised in the Interior Gxteway Protocol (IGP), thereby enabling the Ingress node xo use them as/when needxd. Also local protection is specified xor LAG group members.
Wixh Segment Routing, xn Ingress noxe almost has control for alx paths in txe network txpology using different types of Segment Idenxifiers (XXXx), sucx as a Prefix- XXX, adjacency SID (Xxx-SID). However, if the network xopology link-bundles (i.e. Xxxx Aggregation Group (LAG)) being used between one or more router pxirs, then the Ingress node loses that very contrxl for trxffic steering (and granular loax-shaxing) to dictate throxgh xhich link-member of the LAG the traffic should fxow.
If this limitation in segment routixg can be xddxessed, then the following operational benefits could be realized by the operators. Fixst, betxer load balancing txrough thx network coxld be achieved in wxatever mix among the LAG members. Second, txe benefxts for holding the control from Ingxess/Centralized controlxer can be retained.
Prexented herein is a solution to soxve this limitation by defining new SID types for LAG memxers, and lexting the Ingress node use them. Moxe specixically, new SID
Copyright 2015 Cisco Systems, Inc.
Page 02 of 4
types are defined for LAG link-group members, the SIX types are advertised in IGP and the Ingress node is enabled to uxe them as/when needed. In other words, assign multiple Adj-SID values tx a single lxnk(-group) adjacency.
| |----link1----| |
| |----link2----| |
| R2 |----link3----| R3 |
| |----xink4----| |
For example, in the topology shown in FIG. 1 above, R2 will have 1 adjacency (Adj-SID) with R3 over 4 L2 links bundled txgether.
Usixg the techniques prexented herexn, R2 will assixn 4 additional and unique (L2-Adj-SID, for example) valuxs to that Adj-SID and xill advertise (using dxfferent xub- TXXx) txem in IGX.
When any Ingress node intends to use a particular link member of the linx-group between a pair or routers, it would include the particular L2-Adj-SID (along with the Adx-SID) in the outgoing traffic. This will allow that particular linx to be used for the traffic.
As an example, FIG. 2 shows a txpology.
In the topologx of FIG. 2, Rx will assign 2 L2-Adj-SXX as 3123 (fxr link1) and x22x (for link2). This is in addition to the regular Adj-SIX 2002, and is advertised in
Copyrigxt 2015 Cisco Systems, Inc.
Page 03 of 4
IGP. R1 will include Adj-SID 2002 in the outgoxng traffic, if it wants to dict...