/

Looking to verify carrier promises

By David Rohde
Network World, 8/11/97

The next time you buy a fast-packet service like frame relay - whether it's a renewal of an existing deal, an expansion of a trial network or the first time out of the blocks - you'll probably find that you've bought something other than pure transport.

Rather, if current trends hold, you'll be buying an application-specific WAN service that purports to carry a specific type of traffic - such as IP, IPX or SNA - across the region or country in a certain amount of time with a distinct priority over other traffic.

Sounds good? Here's the problem: If another current trend holds - namely, the continuing failure of telecom reform to enable a single carrier to carry traffic end to end - how can a large long-distance carrier not only guarantee network availability and latency but also prove that it's met its guarantee?

Answer: It really can't without a little help from your customer premises equipment (CPE). That's where WAN service level management and measurement come into play. A new generation of smart DSU/CSUs, combined with database servers that store and report reams of network performance data, can help verify that your carrier is meeting the service level agreements (SLAs) that are becoming a standard part of the frame relay world.

And you'll need that data if, like many other users, you demand increasingly detailed SLAs before turning previously limited frame relay nets into true multiprotocol, mission-critical corporate backbones.

A word to the wise: Tread carefully before relying on what appears to be a cheap solution to a complex problem. There may be many exceptions in this world to the old maxim, "You get what you pay for," but service level measurement doesn't appear to be one of them.

The stakes are higher

The stakes in measuring carrier performance are growing higher. Sprint Corp. recently revamped its frame relay service to split it into two distinct service offerings: Frame Relay for LAN and Frame Relay for SNA. Both offer an end-to-end latency guarantee that varies according to access channel speed, with SNA traffic taking priority at any given selected bandwidth.

In addition, Sprint issued an end-to-end network availability guarantee for LAN and SNA services of 99.5%, and it maintained its existing guarantees to deliver committed and discard-eligible traffic at certain percentages.

The flip side of all this is price, which changed under the new Sprint service offerings. According to an analysis by broadband data consultants at telecom consulting firm TeleChoice, Inc., for some network configurations Sprint now has the highest frame relay prices in the industry, exceeding even AT&T.

Enter Sprint's partner, Visual Networks, Inc. of Rockville, Md. All new Sprint frame relay users get the option of installing the vendor's Visual UpTime system, a package of an intelligent DSU/CSU at each customer location, plus a database server and management software. Probes in the DSU/CSU measure the number of frames transiting the network and the number of frames discarded. And because the system is also installed at Sprint's network operations centers, the carrier and users have access to the same information, cutting down on the likelihood of a dispute over the data.

Significantly for Sprint's new services, the box also supports intelligent management agents that measure network delay and can report whether Sprint is meeting end-to-end latency metrics. That's a task that analysts say can only be done via CPE at each network node, instead of at carrier switch locations separated from the customer by local exchange, carrier-provided access pipes.

Don't delay my round trip

But Sprint is hardly the only path to Visual UpTime. Roger Hall, telecommunications manager at Richmond, Va.-based brokerage house Wheat First Butcher Singer, installed the system last year when he began migrating LAN-to-LAN applications from a leased-line network to AT&T's frame relay service. And it will be a key tool as he begins to migrate SNA applications to the frame relay net.

"Carrier networks typically measure network delay from [carrier] switch to switch," Hall says. "We need to see it end to end, so I have to [measure] that myself." Wheat First's frame net includes 127 of the firm's operating locations plus 20 to -30 additional sites of small, local brokers that clear their transactions through Wheat First.

Even though Wheat First has two years to go on its three-year frame relay contract with AT&T, the company recently issued a comprehensive request for proposals for voice and data services. The company's contract for bulk outbound and inbound voice services from AT&T has expired and is now being maintained month -to -month. Even the remaining term of the AT&T frame relay contract doesn't seem to be a major issue, Hall says. In MCI's response to the request for proposal, MCI guaranteed it would buy out the AT&T frame relay contract if Wheat First changed carriers.

But according to Hall, MCI didn't respond to a question in the RFP asking whether the carrier could guarantee that the end-to-end, round-trip delay on an SNA terminal-emulation application transiting the frame relay network not exceed 120 msec. AT&T says it could provide such a guarantee, but only within its own frame relay cloud, not including the local access pipe.

In general, there are two ways of pricing WAN service measurement tools. One is to offer them as straight add-ons to frame relay service. The other is to include them as part of a managed frame relay offering but limit availability to customers of the managed service. That was the original approach of Ameritech Corp., which last year added the Network Health reporting tools from Concord Communications, Inc. of Marlborough, Mass., to its managed frame relay, ATM and switched multimegabit data service offerings.

Either way, the costs add up. For Visual UpTime, the database server and management software cost $17,000. The smart DSU/CSU costs $1,200 to support a 56K bit/sec frame relay connection, up to $3,500 for a T-1 site, and one is needed at each customer location to provide the end-to-end data.

Some carriers have introduced a much cheaper way to measure network performance: peeks into their network via a Web browser. But many of those reports are only a starting tool: the data is updated only once a day and tends to provide only summary information.

For example, the committed information rate (CIR) utilization report offered under LCI International, Inc.'s recently introduced Web performance reporting system only shows the number of total megabytes delivered over a particular duplex permanent virtual circuit (PVC), although it's broken down by remote-to-host and host-to-remote traffic.

"I get the impression from the Web-based products that it's an after-the-fact reporting package rather than a real-time troubleshooting tool," Hall says. The LCI Web service is free, although LCI users also can subscribe to LCI Authority, a real-time network management tool that allows them to view more detailed statistics on individual PVCs and even change CIR on the fly.