New challenges, New choices

Frame relay links pose unique testing challenges. Luckily, technicians can use a broad array of portable test devices that have a wide range of capabilities

Seeking to save money, many enterprise network managers are turning to frame relay for their data communications needs. And integral to frame relay's reduced cost is bandwidth sharing among customers within the service provider network.

As the demand for frame relay has grown (Figure 1), customers also have begun to demand assurances that they are receiving the bandwidth they are paying for. Service providers have responded by offering service level agreements, which attempt to define quality of service (QOS) in terms such as committed information rates (CIRs) and maximum errored seconds.

This, in turn, has created an increased need for portable frame relay test equipment-both for installing new circuits and for ongoing service monitoring. The test equipment industry has responded by adding new functions to existing systems and, in some cases, by developing new products.

Frame relay basics Frame relay is a fast-packet, link-level protocol. Most of what's different about frame relay involves establishing links between nodes, including the local link between the user and the network point of presence: the frame relay switch.

Once frame relay packets are received at the switch, they are transported through the network in much the same way as any other packet. As a result, there is much specialization in the installation tools used to establish the local links and very little differentiation among the protocol analyzer tools used to monitor packet transmission within the network.

Frame relay local loops are implemented over digital access circuits. Therefore, testing a new line requires testing both the digital circuit and the frame relay service. Both are checked by setting up a "conversation" between the test equipment and the switch to which it is attached. The test equipment sends "request for status" messages, and the host responds.

The digital access link is at the physical level (Layer 1 in the international Open Systems Interconnection model). Checking a number of factors, including circuit status, data rate, service codes and circuit quality, determines whether the access link is ready for use.

Circuit status checks whether the circuit is complete and asks if it can transmit a data mark.

The central office switch and the customer premises equipment-in this case the test equipment-must be able to match clock rates. The data rate is the maximum clock rate on which the two elements can synchronize.

The network can transmit service codes to the customer premises. These codes can represent loopback commands or alarms indicating network conditions, which include latching/non-latching channel service unit loopbacks; out-of-service conditions, when the far end of the digital circuit is not transmitting; red alarms, when the far end of the network is not transmitting; yellow alarms transmitted from a device receiving a red alarm; and blue alarms, an "all ones" pattern that maintains synchronization for a device at one end of an incomplete circuit. The network uses latching/non-latching loopbacks for both subrate and T-1 links; out-of-service alarms only on sub-rate links; and red, yellow and blue alarms for T-1 links only.

Circuit quality looks at subrate and T-1 circuit errors such as bipolar violations, frame slips, loss of carrier, cyclic redundancy check errors and the number of seconds in which errors occurred (errored seconds). Frame slips occur for T-1 links only.

What we call frame relay is the link-level protocol-the lower half of Layer 2 in the OSI model. The frame relay connection to the network is confirmed by checking continuity, link management type, frame relay packet source, data link connection identifiers (DLCIs) and CIRs.

A digital circuit must be terminated in a frame relay switch in the CO. A continuity test verifies that termination.

Three types of link management are currently used on permanent virtual circuits in frame relay networks: LMI (Annex A), Annex D and X.36. The CPE and the network must use the same link management type.

The frame relay packet source consists of three interfaces. The user-to-network interface (UNI) is on the user side of the user-to-switch link. The network interface is on the switch side of the link. And the network-to-network interface (NNI) is within the carrier network.

The test equipment signals that it is a UNI. The switch should respond with a "network interface" confirmation.

DLCIs are identification numbers assigned to the local loop ends of virtual circuits and are of local significance only (Figure 2). However, for the first leg of a frame relay circuit to exist, the DLCIs between the UNI and the network interface must match. DLCI status may be active, inactive, new or deleted.

The CIR for each DLCI is determined in advance. The rates are unidirectional and may be different in each direction on a bidirectional circuit.

Finally, the amount of bandwidth available beyond the CIR is network-dependent and varies. The CIR measure is of special interest in defining and measuring QOS.

Packet-level testing Frame relay is a packet protocol, meaning that the information payload being transmitted is wrapped in a packet between certain header and trailer bit strings.

The factors listed above for data access circuit testing and for frame relay link testing can be checked without opening a packet. Some convenient or important measures cannot be tested in that manner, however. These include flow control, discard eligibility, sequence and payload.

Flow control is when two header indicators in management packets are used to notify users of network congestion-forward explicit congestion notification and backward explicit congestion notification. Backward explicit warnings are sent to nodes that have sent packets. Forward explicit warnings go to nodes that receive packets.

Discard-eligible header bits are used by the frame relay network to identify frames that represent traffic that exceeds the committed rate. Discard-eligible frames may be dropped at any point in the network to control congestion or manage throughput.

In addition, frames may be assigned sequence numbers so that recovery is possible if traffic is discarded.

Some network problems can be diagnosed only by examining the data content of the packet. For example, a node expecting a TCP/IP packet may actually receive X.25.

For installing frame relay circuits, technicians generally use either a small hand-held frame relay tester, such as the Looker or Looker HS from HT Communications, or a somewhat larger, portable frame relay tester, such as the Frame Tester F4100 from Fastcomm. Such devices are designed specifically for frame relay testing and are relatively inexpensive-$2000 or less.

Ideally, such testers should address the problem of too much demand and too few technicians. HT's products, for example, are almost completely passive devices that look at but don't touch configuration settings. They were designed with the idea that a mailroom clerk or secretary could read the contents of the display to a technician over the phone, if necessary.

Another approach, used by Fastcomm, is to enable a portable tester to be used by a local operator or interrogated from a host site.

Once the local links have been established and their operating parameters confirmed, further testing requires more sophisticated tools that are capable of opening and decoding packets, dynamically measuring network performance over time to satisfy service commitments, and troubleshooting network components.

Since they operate at the packet level and since frame relay packets are treated much like other packets, it should be no surprise that these more sophisticated tools are not frame relay-specific. For most of them, frame relay capabilities are a software extension-in fact, some of them claim compatibility with hundreds of protocols.

The simplest of these are simulators that play the role of terminals or other user equipment that move data into the network, or simulate traffic within the network for testing internal components. Fastcomm and Digitech offer portable products with this capability.

Performing more sophisticated testing than simple traffic generation requires additional intelligence. Test equipment manufacturers have responded in two ways: by developing equipment or software that requires a laptop or other PC for operation, and by developing systems that include their own displays and computing power. In most cases in which a PC or equivalent is not included, a board-level component is. Companies offering frame relay test devices with an extra level of intelligence include Digitech, Frederick Engineering, Radcom and Wandel & Goltermann. These devices cost from about $7000 to more than $10,000.

At the high end of portable frame relay test devices are stand-alone protocol analyzers. Vendors addressing this niche include Hewlett-Packard, Network Communications, Network General, Tekelec, Telecommunications Techniques Corp. and Wandel & Goltermann. These products sell for $15,000 to more than $30,000.

Different test tools are appropriate for different needs. Installing new frame relay circuits, for example, requires only the simplest test tools. The tools are pre-programmed and the installer doesn't need an engineering degree to use them. Prices are low, and the return on investment can be very high because installations can be done in 20 minutes if all goes well. Without frame relay-specific tools, installation problems typically take two days and 20 minutes to resolve-two days to define the problem and 20 minutes to fix it.

However, simple installation tools are not much help in diagnosing problems beyond the link level. For example, frame relay headers contain the FECN and BECN bits, which can't be checked unless the test equipment can decode and read the packets. Some installation-only tools lack that capability.

In addition, frame relay is a rather merciless protocol. If a user sends more traffic than a service provider has guaranteed to transport, the excess frames will be given discard-eligible flags and may be discarded by the network. Technicians need to be able to read packets to check frame sequence numbers to see if that has happened.

What about the contents of the information field of the frame relay packet? If an X.25 packet has been encapsulated and the receiver is expecting an IPX message, the only way to determine the problem is to decode another layer. That requires more intelligence. And if it's necessary to compare what a switch or a router is receiving with what it is transmitting, the user will need a dual-port analyzer.

Generally speaking, the more layers that have to be decoded or the faster the tool has to operate, the more expensive the tool will be. Equally important, the more difficult the problem, the more training the technician will require to operate the test equipment. Those are two good reasons for test equipment purchasers to buy only what they need for a particular type of application.