Provisioning Frame Relay


Provisioning Frame Relay.  The following table summarizes the major steps involved in provisioning a frame relay network.




Step 1 Choose a committed Information Rate Based on Realistic, Anticipated Traffic Returns
  • Determine how much interactive, file transfer and broadcast traffic the link must support.
  • Use some sort of traffic shaping features, like the Cisco features to be outlined later to improve predictability.
  • If you opt for an inexpensive zero CIR - Committed Information Rate, try to get a service level agreement from your provider
Step 2 Aggregate All CIRs to Determine Core Bandwidth Requirements
  • Add up the CIR at each access site.
  • If the access site has a zero CIR, use half of the link speed as a rough estimate.
Step 3 Determine the Link Speed and Number of Interfaces Required on the Core Router
  • Determine the required number of interfaces and the number of DLCIs - data link connections identifiers each interface should handle.
  • If there are multiple protocols, then decrease the number of DLCIs per interface.
  • If an application uses many broadcasts, consider static routes and SAP filtering.
Step 4 Choose a Router Platform that can Handle the Job
  • Consider
    • processing requirements
    • memory requirements
    • interface density
  • For example, if you have 10 access sites, each provisioned with a 32 kbps CIR, you will need 320 kbps bandwidth.  You can use a router with one T1/E1 port utilizing ten DLCIs.


Traffic Shaping Over Frame Relay.  While this presentation primarily focuses on features available in Cisco devices, it still represents a good survey of traffic shaping features you are going to want to consider.

Generally, Cisco's Frame Relay Traffic Shaping feature eliminates bottlenecks by providing high speed connections at the central site, with lower speed connections at the branch sites.  In order to manage network congestion Cisco defines three performance measures.

  • CIR - Committed Information Rate
  • FECN/BECN - Forward/Backward Explicit Congestion Notification
  • DE - Discard Eligibility bit

The Frame Relay traffic shaping features should provide the following capabilities or some variant to improve scalability and performance.


Feature Description
Rate Enforcement on a per VC Basis You can configure a peak rate to limit outbound traffic either to the CIR or to some other defined value, such as EIR - excess information rate.  Use rate enforcement to
  • Limit the rate at which data is sent to the VC at the central site.
  • Improve performance in conjunction with the existing DLCI prioritization feature.
  • Allow other criteria, such as CIR, to control the router's transmission speed.
  • Preallocate bandwidth to each VC, creating a virtual time-division multiplexing network.
Generalized BECN Support on a per VC Basis The router can monitor BECNs and throttle traffic based on BECN marked packet feedback from the Frame Relay network.  The traffic is dynamically throttled based on information contained in BECN tagged packets.

BECN based throttling is done by holding packets in the router's buffers to reduce data flow from the router to the Frame Relay network.  This is done on a per VC basis and the transmission rate is adjusted based on the number of BECN tagged packets being received.

PQ/CQ Support at a VC Level The PQ/CQ Support feature at the VC level allows for fine granularity in the prioritization and queueing of traffic.  These features improve scalability and performance by increasing the density of virtual circuits and improving response time.  This applies to the  Frame Relay PVCs and SVCs.

Custom queueing with per VC queueing and rate enforcement capabilities enable Frame Relay VCs to be configured to carry multiple traffic types (such as IP, SNA, IPX) with bandwidth guaranteed for each traffic type.


All of these features require the router to buffer packets to control traffic flow and compute data rate tables.  Because of this, router memory and CPU utilization must be adequate and configured so that the additional work to implement the features does not degrade overall Frame Relay performance.