x
Optical/IP

VPLS Standard Debated

Virtual private LAN services, which extend the corporate Ethernet network over a carrier’s wide area network, are expected to gain momentum among service providers in 2003. But vendors will first have to get behind one of several Internet Engineering Task Force (IETF) standards drafts before service providers will adopt the technology in any big way.

“It’s important for vendors to rally around some draft of a standard,” says Kevin Mitchell an analyst with market research firm Infonetics Research Inc. “Today’s carrier networks are multivendor, and interoperability is very important. Even with standards in place, that isn’t easy to achieve, but it’s a start.”

According to a recent Infonetics study, 32 percent of carriers are already offering virtual private LAN services. By the end of next year, 47 percent say they expect to be offering them. The study, called Service Provider Networks: Access, Routing, Switching, and Optical, US/Canada 2002 asked 20 carriers which access technologies they are deploying now or will deploy in the next 12 months.

“It’s a positive change, but not overwhelming,” says Mitchell about the results. He says the uncertainty in the standards process could be one reason that more carriers are not bullish on deployments of VPLS for next year.

But the study found that carriers are deploying and plan to deploy more simple point-to-point Layer 2 Multiprotocol Label Switching (MPLS) Virtual Private Networks (VPNs). These are VPNs that use Frame Relay or Asynchronous Transfer Mode (ATM) as transport, instead of the Ethernet used in VPLS. Fifty-three percent say they offer that service today, and 74 percent say they expect to be offering it by late 2003.

These results aren’t too surprising, on two fronts. For one, many carriers plan to continue leveraging their existing Frame Relay and ATM networks. And secondly, the standards process is more mature for point-to-point or circuit-based Layer 2 MPLS VPNs. There are currently two IETF drafts to support these circuit-based VPNs (see Juniper/Cisco Duke It Out Over MPLS). These include the "Martini Draft," named after its author Luca Martini of Level 3 Communications Inc. (Nasdaq: LVLT), which uses Label Distribution Protocol (LDP) signaling to establish point-to-point connections throughout the network.

The other Layer 2 VPN draft comes from Kireeti Kompella, an engineer at Juniper Networks Inc. (Nasdaq: JNPR). This draft uses the Border Gateway Protocol (BGP) common in routed networks for signaling. While Juniper claims that Draft Kompella scales better than Martini, most vendors have standardized on Martini.

The standards situation is a bit stickier when it comes to VPLS. Because Ethernet by nature is not point-to-point, Draft Martini alone is not sufficient. Currently, there are several IETF drafts being debated in the Provider Provisioned Virtual Private Networks working group. Over the past several months, vendors have begun choosing sides in the standards face-off, and it appears that most are rallying around one of two proposals: The "Lasserre-V. Kompella" draft, which extends the use of LDP for signaling defined in Draft Martini; or the original Kompella Draft proposed by Juniper.

Marc Lasserre of Riverstone Networks Inc. (Nasdaq: RSTN) and Vach Kompella of Timetra Networks (who happens to be the brother of Juniper engineer Kireeti Kompella) are the two main authors of Lasserre-V. Kompella. So far, the draft has gained official support from engineers at: Other vendors like Foundry Networks Inc. (Nasdaq: FDRY) are already supporting features in the Lasserre-V. Kompella draft, says Chandra Kopparapu from Foundry. The company demonstrated its capabilities in an interoperability demonstration at the Networld + Interop show in Atlanta back in September (see Foundry Takes Part in VPLS Test).

“The Lasserre-V. Kompella draft falls more easily around current Ethernet switching technology,” says Paul Joseph, a product manager at Force10 Networks Inc.. “Companies like us, Cisco, Extreme, Riverstone, Foundry, Timetra -- we are all leaning towards the Lasserre-V. Kompella draft, because we have no problem manipulating Ethernet packets.”

The other draft proposed for VPLS is the original Kompella Draft from Juniper.

“The sweet spot for Layer 2 MPLS VPNs deployments seems to be in using them to complement Layer 3 MPLS VPNs. Like RFC 2547, Kompella also uses BGP, so it’s an easy fit,” says Kevin Dillon, director of portfolio marketing at Juniper.

Dillon says that LDP requires much more manual configuration from service providers when adding new nodes to the network. More manual configuration means more cost, he notes. Using BGP is simpler.

“That is true,” says Force10's Joseph. “But a lot of service providers are nervous about deploying BGP because the protocol itself is so complicated. They’d rather use a simpler protocol even if it means more manual intervention.”

Technology aside, the debate will be resolved by politics and market share. The Lasserre-V. Kompella draft already has two things working in its favor: It has support from incumbent carriers like SBC, which has three engineers officially supporting the draft; and it also has support from Cisco, the number one player in the Ethernet switching market. With endorsements like these, it’s difficult to bet against it.

— Marguerite Reardon, Senior Editor, Light Reading
www.lightreading.com Want to know more? The big cheeses of the optical networking industry will be discussing service trends at Lightspeed Europe. Check it out at Lightspeed Europe 02.

CarrierClass 12/4/2012 | 9:16:23 PM
re: VPLS Standard Debated Following LRGăÍs informative and accurate VPN report by Geoff Bennett/Daniel Proch last month, I find it extremely disappointing that they have gone on to publish such an inaccurate article.

LR: According to a recent Infonetics study, 32 percent of carriers are already offering virtual private LAN services.

By Virtual private LAN services are you referring to the IETF VPLS drafts? If so, which vendors are these carriers using then? As far as I am aware only Juniper, Riverstone and Foundry have any VPLS capabilities and these are not mature yet. So... are you saying that 32% of carriers have deployed MPLS networks with all Juniper/Foundry/Riverstone PE's and are offering customers VPLS services over networks running alpha/beta code??

LR: But the study found that carriers are deploying and plan to deploy more simple point-to-point Layer 2 Multiprotocol Label Switching (MPLS) Virtual Private Networks (VPNs). These are VPNs that use Frame Relay or Asynchronous Transfer Mode (ATM) as transport, instead of the Ethernet used in VPLS.

Point-to-point Layer 2 MPLS VPNs can be used to transport ATM, FR, AND Ethernet G㢠not instead of Ethernet.

LR: Fifty-three percent say they offer that service today, and 74 percent say they expect to be offering it by late 2003.

WHAT?? So 53% of carriers offer ATM and FR services over an MPLS network right now? Which carriers are these then and how do they price their ATM/FR *Like* services over MPLS compared with their existing ATM/FR services over their ATM/FR infrastructure?

FR: These results arenGăÍt too surprising, on two fronts. For one, many carriers plan to continue leveraging their existing Frame Relay and ATM networks.

Well if they are offering ATM/FR services over MPLS how are they leveraging their existing FR and ATM networks? Unless you are suggesting that they are running MPLS over ATM and then offering ATM/FR over MPLS??? Or unless you mean they are using ATM/FR access links at the edge over an MPLS core probably running over SONET G㢠But is reusing ATM/FR access links the same as Gă leveraging their existing Frame Relay and ATM networksGăÍ?

LR: The standards situation is a bit stickier when it comes to VPLS. Because Ethernet by nature is not point-to-point, Draft Martini alone is not sufficient.

Draft martini alone is not sufficient for what? There are two Martini drafts, one that deals with encapsulation and one that deals with transport. The VPLS drafts use the Martini draft for encapsulation anyway. If you are referring to the Martini transport draft what is it not sufficient at doing? It is more than sufficient for tunneling Ethernet traffic across an MPLS network using point-to-point links. Using this method the customer premises equipment is responsible for MAC address learning and it has the same scaling problem of any point-to-point topology where n sites requires ([n] [n G㢠1]) / 2 links for full mesh connectivity. If a customer has a large number of sites and requires full mesh connectivity then yes VPLS is a more sensible approach.

Keeping up to date with whatGăÍs going on with all the IETF drafts is difficult enough without inaccurate reports like this.

Very Disappointing.

CC
gigeguy 12/4/2012 | 9:15:32 PM
re: VPLS Standard Debated I'm not a marketing guy, so I can't comment on the market share numbers. However, Marguerite accurately described the popularity of the Lasserre-VKompella draft for VPLS in the IETF.
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE