WAN Design Takes an Applications-Focused Approach

Khali HendersonBusinesses’ increasing reliance on hosted and cloud-based applications is making choosing the right WAN technology even more critical. Channel partners, then must resist the temptation to simply respond to a typical request for quote for an IP MPLS VPN with just a price. Specific customer applications and quality of service requirements may call for a different WAN technology like Ethernet, VPLS or even a hybrid approach.

Giving the customer what they want can backfire as one agent discovered the hard way. On the customer’s request, his company quoted and subsequently won a $30,000 a month deal for a multisite international IP MPLS-based WAN. Not long after, the client began having issues with its ERP system; specifically, it was timing out during the data replication process between two sites. The agent and its carrier sales engineers resolved the issue with an Ethernet private line between the two sites. However, the agent said the incident gave his company a black eye, which could have been avoided if its sales team was armed with more information about the customer’s applications and their use.

“The real debate that’s going on is how do you know when you need Layer 3 MPLS or Layer 2 VPLS,” said Ray Watson, director of sales engineering for Masergy Communications Inc. “You don’t make decisions about Layer 2 or Layer 3 or even Layer 1 based on Layers 2, 3 or 1, but based on Layer 7. It has to be based on what the application is.”

Watson is referring to the Open Systems Interconnect (OSI) model, which has seven layers, starting with Layer 1, which describes the physical layer and ascending to Layer 7, or the applications layer. Layer 2, where Ethernet falls, is the data link layer, which provides error-free transfer of data frames from one node to another over the physical layer. Layer 3 is the network layer, which  controls the operation of the subnet, deciding which physical path the data should take based on network conditions, priority of service and other factors. Layer 3 is the known as the IP layer. Technically, MPLS, or multiprotocol label switching, operates essentially in between Layers 2 and 3 and can overlay existing technologies such as ATM or Frame Relay, or operate in an entirely IP native environment. That said, most references to “MPLS networks” assume an IP-based MPLS VPN.

Assess Customer Applications

Watson advises channel partners to get their customers thinking in terms of Layer 7 applications in order to make a recommendation for a WAN design.

“I would rather know what they are doing with BYOD and cloud computing than how many T1s they have. That will have an impact on the applications and down the [OSI] stack,” Watson said. So, for example, he wants to know if they have deployed mobile device management (MDM), and are integrating Microsoft Lync with Avaya or trying to replicate its ERP system across data centers. “If they don’t know, then we will start gathering that information. We are more interested in what’s coming than what’s in the ground.”

Heather Selbert, vice president operations for network reseller American Telesis, said understanding the requirements of each application is the next step. If the customer is replicating ERP data between two sites, for example, it’s important to know how sensitive the application is to latency. If the customer doesn’t know, they can go to their application provider and get the information needed. If there is a low tolerance for latency, then a routed IP MPLS network may not work and a private line may be better, she explained.

This data synchronization requirement is a big red flag for WAN designers choosing between Layer 2 and Layer 3 technologies. Watson explained that with Layer 2 VPLS, all locations are on the same subnet and function as if they are on the same LAN. With IP MPLS, the locations are on separate subnets and data must be routed across them, increasing latency.

Quality of service (QoS) requirements for voice and video are another one. “Customers are definitely embracing VoIP between locations, so being able to say [to them], ‘I’ve got the class of services with quality of service on my MPLS network,'” they like that,”  said John O’Hara, principal with SE Solve.

MPLS is known for its traffic engineering, failure protection and the ability to guarantee QoS over IP.

Selbert said VPLS can provide some similar  classes of service through the use of VLANs, which she describes as a “mini network in your larger network.” “You can segment your network into separate pieces so they don’t steal bandwidth from each other,” she said.

In order to get information about customers’ application requirements, Selbert suggests agents ask customers simple questions such as:

  • How many sites do you have?
  • Do the sites have to talk to each other?
  • What applications are you running or planning?
  • Do you need to replicate data between sites?
  • Do you need to put voice or video on the same paths that handle replication?

Taking Stock of Customer IT Resources

A second area to consider is the customer’s IT resources, specifically their ability and desire to control their own routing. “That’s a two-phased question,” said Watson. “You have to have the folks that can understand routing and you also have to want to do it.”

To oversimplify, VPLS gives customers complete control of their network while control of the IP MPLS network is shared with the carrier.

“The customer’s IT guys know how to run a LAN. They know they can segment their traffic and do voice and video and data and best-effort Internet on their LAN. They know how to do VLANs. They want that to happen in their WAN too,” explained Selbert, noting that VPLS gives them that ability. “MPLS can do that, but it’s really hard to configure.”

Specifically, the customer has to set up IP addresses and every time they want to change something, they have to reconfigure it in their router and also call the carrier to configure its routers, Selbert said.

That said, there is a growing demand for managed IP MPLS or managed router, wherein the carrier manages the customer and carrier routing.

“A lot of providers offer managed MPLS,” said SE Solve’s O’Hara. “The customer is only responsible for the Ethernet port into their network. That’s pretty nice. I have a lot of customers embracing that.”

In many cases, managed IP MPLS customers are smaller or with few IT resources. But that’s not always the case. O’Hara describes one 11-location customer with  IT-savvy staff that opted for the managed IP MPLS because it didn’t want to expand to handle this function. In addition, the provider gave them the visibility they needed into their network usage through Web-based software that would have cost them hundreds of thousands of dollars, an expenditure they previously were considering. Including that capability in with the monthly payment sealed the deal.

Even for customers that can and want to manage their own routing, change management policies can favor one technology over another, said Selbert. She explained that many Wall Street firms, for example, have rapid fire change management, meaning they have a narrow window of opportunity to make changes to their network like 10 p.m. on a Friday. “If they are going to get hung up waiting on the carrier, then I don’t want to propose MPLS,” she said.

Considering Cost

While there are many considerations in designing a WAN, there is no reality check greater than the customer’s budget. “They might want to get the great MPLS network fully managed across 20 locations, but the budget might only support IP VPN,” said O’Hara.

In addition, he added, customers often make assumptions about price based on old information. “Customers might say, ‘I know Ethernet is going to be expensive, so I need to go with MPLS.’ But that might not be true,” O’Hara said. “Ethernet prices have come down dramatically.”

In his recent experience, for example, a customer believed that an Ethernet connection between data centers in Atlanta and Northern California would be too pricey. Instead of assuming they were right, O’Hara’s team decided to compare and found the Ethernet was the better price performance option for the job, which was to move eight petabytes of data from one location to another.

“Agents just can’t assume what the customer is asking for is correct,” he added. “They have to be able to educate the customers on what’s out there, what customers might not know.”


Hear more about Ethernet and IP MPLS in the session, “WAN Design Basics for Distributed Workforces and Cloud Workloads,” at the Channel Partners Conference & Expo, Feb. 27-March 1, in Las Vegas. Or check out the report, “High-Speed Ethernet Services: Channel Opportunities & Challenges,” which is available for download at the Channel Partners Resource Center.

Leave a comment

Your email address will not be published. Required fields are marked *

The ID is: 72692