& cplSiteName &

Kompella Backs BGP

Light Reading
News Analysis
Light Reading
10/28/2003

WASHINGTON, D.C. – Kireeti Kompella, Distinguished Engineer at Juniper Networks Inc. (Nasdaq: JNPR), is on a mission to set the record straight on Border Gateway Protocol.

Today, during the MPLS 2003 conference, he gave a presentation urging his colleagues to take another look at BGP as the signaling and setup protocol for VPLS, rather than extending Label Distribution Protocol (LDP).

Virtual Private LAN Service or VPLS is an emerging standard that creates a point-to-multipoint Ethernet network using Multiprotocol Label Switching (MPLS). The biggest technological debate in the working group is over which signaling protocol to use: BGP or LDP (see VPLS Standard Debated). Both camps have their zealots, and for awhile it looked as if the LDP draft, co-written by Kireeti’s brother Vach, was taking the lead (see Kompella vs Kompella). But the tide seems to be turning.

Kireeti Kompella says that BGP already does many things that carriers want a VPLS network to be able to do, like auto-discovery and provisioning. These are things that would have to be added to LDP.

“BGP has some flaws,” he admits. “You have to tweak this or that. But it’s a much more pragmatic approach than building a new, solve-everything protocol.”

He also seeks to dispel what he feels are misconceptions about BGP -- in particular, that BGP is overloaded and that the protocol is trying to do too much.

“The whole premise of multiservice is to put multiple services on one box,” Kompella says. “It will use the same amount of CPU and memory whether you’re running six protocols or one.”

"I think carriers might be more comfortable with this BGP overload issue if they knew that routing processes like this were implemented in a resilient way -- using totally isolated memory areas, for instance", says Geoff Bennett, chief technologist of Heavy Reading, Light Reading's market research division. "Right now, if BGP goes down, it takes a lot of functions with it and tends to take a long time to recover. Graceful restart will help, but better code would help a lot more."

Many large long-distance carriers are very interested in BGP signaling because they’re already running BGP in their networks, Kompella says. What’s more, many of them are already offering Layer 3 VPN services, so it makes sense to use the same signaling protocol for their Layer 2 VPNs as they would use for their Layer 3 VPNs.

Loa Andersson, co-founder of the consultancy, TLA-group and chair of the Layer 2 virtual private network (VPN) working group in the Internet Engineering Task Force (IETF), says that he hopes both protocols will be a part of the final draft.

“There’s no reason not to progress both approaches,” he says. “BGP is used in deployments where carriers are already using BGP routers. And LDP is used in networks already doing Layer 2 point-to-point VPNs.”

While the standard is still a long way from being finalized, many vendors have already put their money behind LDP signaling. In fact, that’s one reason Isocore, which is sponsoring the conference, chose to use LDP signaling in its live VPLS network demonstration. Boxes from Cisco Systems Inc. (Nasdaq: CSCO), Extreme Networks Inc. (Nasdaq: EXTR), and Alcatel SA (NYSE: ALA; Paris: CGEP:PA) are providing wireless Internet connectivity to attendees at the MPLS 2003 conference via a VPLS network.

“I really wanted to be able to test both LDP and BGP VPLS signaling,” says Rajiv Papneja, manager of validation and product evaluation for Isocore. “But it made more sense to just run LDP, since I’ve been testing it for a year. I know it’s stable, and there were vendors able to donate products.”

But Yakov Rekhter, another Distinguished Engineer at Juniper and the father of BGP, says that these debates are irrelevant.

“You know what they say about the pudding,” he says. “ ‘The proof of the pudding is in the eating.’ In the end, the market will decide.”

"Yakov is absolutely right about the market deciding; that's the classic IETF approach," says Heavy Reading's Bennett. "The problem is that vendors end up dissipating resources in developing and supporting both mechanisms until the market really does give a clear indication. It took CR-LDP a couple of years or more to die, and that's a lot of wasted effort. But, like it or not, that's the way the IETF does things."

— Marguerite Reardon, Senior Editor, Light Reading

(5)  | 
Comment  | 
Print  | 
Newest First  |  Oldest First  |  Threaded View        ADD A COMMENT
cc_junk
cc_junk
12/4/2012 | 11:18:34 PM
re: Kompella Backs BGP
While there can be a strong argument for using BGP for discovery and signaling for VPLS it definitely is a very poor fit for VPWS (the point-to-point "Martini" virtual circuit services).

I think many people often forget the differences between the two types of Layer 2 VPN services.
lr_monger
lr_monger
12/4/2012 | 11:18:31 PM
re: Kompella Backs BGP
Agreed 100% -- the current situation is that VPLS is not broadly deployed, so the pain of configuring a full mesh topology in every router has not been experienced by network operators.

For example, when a site is added to a LDP-based VPLS domain every router needs a configuration change to add the reachability information. :-(

With BGP-based signaling, the BGP route reflector does all the heavy lifting.

lr_monger
boobasan
boobasan
12/4/2012 | 11:18:30 PM
re: Kompella Backs BGP
> For example, when a site is added to a LDP-
> based VPLS domain every router needs a
> configuration change to add the reachability
> information. :-(

but at least you are provisionning very tangible and familiar connectivity objects and not via policies which you would still have to do in BGP... It's not that it will all be just magically discovered, right? You still have to configure a lot of stuff, plus, how would it fit into traffic engineered QoS-aware network?

LDP-based VPLS is a very simple network management problem that can and should be solved as such. I don't see why creation and monitoring full mesh is considered to be such an evil thing.
tmc1
tmc1
12/4/2012 | 11:18:29 PM
re: Kompella Backs BGP
that can and should be solved as such. I don't see why creation and monitoring full mesh is considered to be such an evil thing.
==============================================

boobasan,

YOU are an EVIL thing!!!!!!!!!!
;)

stay away from the cognac and red bull you sexy beast.

love and kisses,
you know who...
go_ON
go_ON
12/4/2012 | 11:15:27 PM
re: Kompella Backs BGP
"the tide is turning" ?

Any backup behind this - who apart from Juniper have KKompella draft support. If no one is supporting this then no customers are asking despite what Juniper say.

BTW I think KKompella is a better approach esp for 2547 deployers.

Featured Video
Upcoming Live Events
October 22, 2019, Los Angeles, CA
November 5, 2019, London, England
November 7, 2019, London, UK
November 14, 2019, Maritim Hotel, Berlin
December 3-5, 2019, Vienna, Austria
December 3, 2019, New York, New York
March 16-18, 2020, Embassy Suites, Denver, Colorado
May 18-20, 2020, Irving Convention Center, Dallas, TX
All Upcoming Live Events
Partner Perspectives - content from our sponsors
Multiband Microwave Provides High Capacity & High Reliability for 5G Transport
By Don Frey, Principal Analyst, Transport & Routing, Ovum
5G + Cloud + AI + Ecosystem, Opening New World of Video
By Samuel Chen, President, Cloud & Data Center Marketing, Huawei
All Partner Perspectives