Understanding and Managing FR Congestion

Not all the sources of loss and delay necessarily lie within the management purview of the enterprise operations staff. While the ops staff has considerable access to individual enterprise network elements and to CPE components of a Frame Relay access line, some of the Frame Relay network - the switching fabric, for example - remains a blind spot. When the ops staff have reached a point where they suspect that the Frame Relay network may be the source of unexpected delay, they toggle the Bottom panel to view Congestion Activity to reveal graphs that illustrate the percentage of seconds where frames received from the network had Congestion Notifications (CN) set.

While FR implementations vary, congestion notifications typically mean that FR switches are receiving frames from subscriber access lines faster than they can process and deliver them, causing frames to be held in buffers longer (increasing delay) or, in the extreme case, discarded. If the congested seconds percentage is high, the Frame Relay switching fabric may be having difficulty meeting service levels.

Frame Relay subscribers can put such information to good use. In practice, both FR user and network provider benefit from a cooperative effort to avoid congestion on the FR network. Properly instrumented router implementations support packet scheduling and traffic shaping mechanisms which react to CN’s by dynamically adjusting operating rates over FR down when congestion is present, and up when congestion abates. Certain router implementations require tuning of these parameters, but absent the kind of concrete data derived from a tool like the Congestion Activity monitor, insufficient data is available to support such fine tuning.

Managing Bursty Data over Frame Relay

A unique feature of the Visual UpTime toolset is the Visual Burst Advisor. This feature displays upstream and downstream Access Channel (FR, HDLC, X.25) or FR PVC utilization in three formats:

• Utilization graphs display throughput as percentage of either Access Channel speed or CIR, sampled in 15 minute intervals with one second granularity across a two hour or two day period. Using this graph, ops staff can easily spot brief traffic bursts as red or orange spikes, indicating utilization greater than 80%.

• Pie charts provide a breakdown of utilization by percentage, allowing ops staff to summarize Access Channel or PVC utilization during 15-minute intervals - this is especially useful as an aggregation tool when traffic flow is highly variable.

• For trending over a longer period, the Burst Advisor dialog box provides bandwidth summarizes and CIR utilization during the past two weeks, and performs a three-tier analysis to produce recommended capacity planning actions for the affected Access Line or PVC: increase, decrease, or no change.

Figure 11 illustrates a 56 Kbps FR access channel at a brokerage location. The graphs show that traffic flowing downstream to the brokerage location (from Network) fluctuates between 0-40% of channel capacity, while upstream traffic (to Network) is a steady-state background hum utilizing less than 10% of the channel. By scrolling these graphs left to right over a two-day period, ops staff can visually spot patterns in traffic flow, and zoom into the two-hour period where utilization was unusually high (e.g., the period surrounding 12:15 where a burst of yellow and orange indicate spikes of up to 90% of this channel's 56 Kbps bandwidth).

However, the pie chart indicates this brief burst represents but a small fraction of overall utilization in the short-run (exceeded 40% utilization less than 3% of the time), and the Burst Advisor verifies a slightly higher utilization trend persists over a longer 2 week period (exceeded 40% utilization less than 7% of the time). Clearly, this FR Access Channel has plenty of capacity to handle growth, and the Burst Advisor recommends that capacity could be decreased to reduce cost without adverse affects, even when employing a conservative planning strategy.

Figure 11. Access Channel Burst Statistics Window

The subject company is comfortable maintaining the currently subscribed CIR. The Burst Advisor merely confirms that a bandwidth "comfort zone" exists at this site, which is reflective of the majority of brokerage locations. As a provider of time-critical information services, the subject company is justifiably conservative. However, it is easy to envision how the Burst Advisor will contribute to WAN bandwidth purchase for this company over time. The graph does allow us to illustrate how the Burst Advisor’s ability to identify under-utilized WAN access lines can help enterprises reduce overall WAN services expenditures.

A two-week histogram may reveal a trend or an anomalous situation. Network planners will invariably find Burst Advisor Monthly Reports the best way to evaluate long-term trends by examining how efficiently bandwidth is used, and can make informed decisions before requesting provisioning changes to Frame Relay service.

Figure 12. Consecutive Advisor Monthly Reports

next...