December 1999
Frame Relay Installation: The Whys And Wherefores
BY JOHN GILES
Although originally thought of mainly as a LAN-to-LAN data connection, frame relay
networks are increasingly used to carry both data and voice traffic, and the voice market
represents an area of significant potential growth for frame relay in the next year. As it
is, it is estimated that between 1998 and 1999 the number of installed frame relay ports
worldwide will increase from 425,000 to 790,000. Of course, with all of these new
installations and potential new uses, there are many challenges. A traditional frame relay
service turn-up involves the installation of CPE at a remote site. The CPE is then hooked
up to a frame relay circuit that is connected to the corporate headquarters at the
opposite end. This article discusses challenges associated with turning up new frame relay
service and how to ensure new service is working properly and reliably.
A USER-TO-NETWORK INTERFACE
The CPE consists of a router or FRAD and a CSU/DSU. The FRAD interfaces to the
Ethernet LAN and sends any data destined for a remote site out a serial port to the
CSU/DSU. The data is then converted to a digital T1 signal for transmission to a frame
relay edge switch. Finally, the edge or ingress switch relays the frames through the
backbone network to the far-end edge or egress switch that is connected to the corporate
office CPE.
At installation time, it is critical to examine the segment of the network path between
the CPE and the ingress frame relay switch. A T1 or fractional T1 transmission cable is
installed by a LEC technician between the nearest central office and the customer site. It
is terminated at an NIU that electrically isolates the outside T1 wires from the CPE in
the computer room. The service provider then connects the T1 from the central office to
the ingress switch and configures the port.
GREMLINS IN THE CIRCUIT
When LEC technicians install a T1 line, they usually perform an acceptance test prior to
handing the T1 over to the customer. A loopback is set up at one end of the T1 circuit and
a BERT is performed using a test set to transmit a unique bit pattern. The transmitted
bits are looped back to the test set's receiver and compared to the transmitted pattern. A
bit error occurs if any bits representing a "one" become a "zero" or
vice versa. This test is usually performed over a period of 15 minutes to establish a rate
at which bit errors occur (see sidebar).
What Can Go Wrong With The T1?
- The T1 is dead because it was never connected at the central office.
- The number of fractional timeslots is not set correctly at the central office.
- The T1 signal is weak or attenuated.
- A BERT was never performed and the circuit has poor performance.
What You Can Do About It
Connect the tester to the NIU on the protected side of the T1 line. This is
called a “terminated” connection because you are terminating the end of the T1
with the tester’s transmitter and receiver. Monitor the signal received from the
central office to ensure the signal is present and strong and that no errors or alarms are
present. This tells you that at least you have a valid T1 connected at both ends and there
are no obvious problems. If errors or alarms are present, alert the LEC and tell them
specifically what is wrong. This additional information validates your trouble ticket and
takes hours off correcting the problem.
Next, have the LEC loop back their end of the T1 circuit for you. Once the loopback is
in place, transmit a BERT pattern (the QRSS pattern best simulates live data) from the
tester at the remote site to the central office where it is looped back to your receiver.
Monitor the number of bit errors that occur over time (see sidebar).
Do not accept the T1 line from the LEC until it can reliably pass the BERT criteria.
Frame relay uses no error correction scheme, so every bit error that occurs on the copper
local loop means that the end station must retransmit the entire frame. This can
noticeably degrade performance.
KNUCKLEHEADS IN THE NOC
There are several steps that must be manually performed by a service provider’s
personnel to establish a frame relay link between two network end points.
Miscommunication, human error, incorrect circuit information, and heavy work loads result
in errors when provisioning PVCs between customer sites.
What Can Go Wrong When Service Is Provisioned?
- The T1 physical line was never connected to the frame relay ingress switch port or is
connected to the wrong port.
- The LMI protocol configured in the switch does not match the CPE configuration.
- The DLCIs for each PVC have either not been programmed into the switch or have been
programmed incorrectly.
- The switch at the far end of the PVC has either not been configured or has been
configured incorrectly.
- The edge-to-edge PVC path through the backbone network either has not been configured or
has been configured incorrectly.
What You Can Do About It
With the tester connected to the NIU, you can set a frame relay tester into a
mode that emulates the CPE. That is, it takes over the functions of the CSU/DSU and the
router/FRAD. In this mode, the tester can transmit frames to and receive frames from the
ingress switch.
The LMI protocol is simply a keep-alive polling session between the CPE and the switch
that occurs on a special DLCI address (0 or 1023). Usually, every 10 seconds a status
inquiry frame is transmitted from the CPE to the switch, and the switch responds with a
status message frame. Every minute, the CPE transmits a full status inquiry to the switch
to get the status of all the configured DLCIs.
A tester will engage the switch in this LMI polling session and report the status of
the configured DLCIs, allowing you to verify the physical connection to a switch port,
that the correct LMI protocol type is configured, and that the right DLCIs have been
entered into the switch and the link can reliably maintain the LMI “heartbeat.”
Also, each configured DLCI should read “Active” if there is operational CPE on
the far end of the PVC. If it reads “Inactive,” then either the PVC path through
the backbone network was never set up or the link is down on the far end.
Once the LMI is configured and working properly, you can check the connection to the
far end of the PVC by transmitting an IP PING packet from your tester to a router or
workstation destination IP address. If this fails, then the edge-to-edge PVC path may be
incorrect or was never established. If it does work, check the round trip delay time for
the PING packet; it should never exceed 300 ms.
“THIS FLIGHT IS OVERSOLD”
One of the reasons a frame relay PVC between two sites is less expensive than a
leased T1 line is because the available network bandwidth is shared. IP data traffic is
bursty by nature, so usually you only need the bandwidth for short periods of time. The
provider’s edge switches then multiplexes several customers together to share the
backbone bandwidth. The carrier may be tempted to oversell the available bandwidth, hoping
that all of the subscribers won’t transmit at the same time. If the switch buffers
overflow, frames are discarded.
To guarantee a subscriber will always get a minimum amount of bandwidth, specify a CIR
in Kbps for each purchased PVC. This rate will fall somewhere between 0 Kbps and the
access rate of the local loop (1,536 Kbps for a full T1). The larger the CIR value, the
more the subscriber pays. Any data that exceeds the CIR is subject to discard if the
switch gets congested.
Why Aren’t You Getting The Bandwidth You Paid For?
- The service provider didn’t program the correct CIR values.
- There is congestion somewhere else in the network.
- The bandwidth at the edge switches is oversold.
What You Can Do About It
You can verify that a new PVC will support a sustained traffic load equal to the
negotiated CIR by using the traffic generator function of a frame relay testeR.
The router on the far end of the PVC need not be taken out of service for the test.
Just configure the far end router (using console commands) to loop back the PVC you are
testing. Then set the transmission rate of the tester to match the CIR that was negotiated
for the PVC and start generating traffic. (Note: The tester must be able to maintain the
LMI keep-alive with the switch at the same time that it transmits test traffic, otherwise
the UNI link will go down and traffic will not get past the ingress switch. Most frame
relay testers have this capability.) Look for:
- Lost Frames: A result of congestion or the frames are corrupted by transmission
errors along the way.
- Throughput Measurement < The CIR: There is a bottleneck along the PVC path. Data may
be buffered for too long in any of the switches along the route. Also an indication of
frame loss.
- Excessive delay: The ITU is in the process of standardizing the maximum one-way
delay allowed edge to edge. Current recommendations for a frame of 576 bytes in length
that travels over a full T1 link is 52 ms nationally and 166 ms internationally. With
round-trip delay values, just divide the values by 2 to get the one-way value.
- Congestion: If a switch port gets congested, it will begin “marking”
the FECN bit in the frame relay header in frames traveling toward the congestion. Also, it
begins “marking” the BECN bit if frames travel in the opposite direction through
the congested port. If the tester reports test frames contain FECN or BECN notification,
then there is a bottleneck.
CONCLUSION
Installers responsible for turning up new frame relay service will attest that
numerous things can go wrong. Both time and finger-pointing can be saved by pre-qualifying
the circuit. If problems are found, you can give the responsible party intelligent,
detailed information about the errors discovered so that they can correct the problem
while you are installing the CPE gear. Pre-qualifying frame relay circuits also builds
your credibility with your service organizations and encourages them to respond quickly to
your service requests.
John M. Giles is R&D network manager, Fluke Networks Division, Fluke
Corporation. As R&D manager, Giles researches emerging LAN and WAN network
technologies and is responsible for the design and maintenance of the division’s
research and development network. Fluke manufactures, distributes and services electronic
test tools. For more information, please visit their Web site at www.fluke.com.
|
TMCnet LOGIN
Webinars
