Model Features: Frame Relay

This section discusses the implementation choices made in developing this Frame Relay model. Certain parts of the protocol have been simplified or omitted in view of the fact that it is intended primarily for performance estimation.

The following information is important in order to determine whether the Frame Relay example model is applicable for a particular simulation study.

The main limitations of the described model are:

  • The model does not incorporate the SVC implementation for call establishment; rather all the virtual circuits used for data transfer are modeled using the PVC approach. The PVC approach assumes that the user sets up the virtual circuits at the service subscription time and these virtual circuits remain intact during the length of the simulation. Specific instructions for setting up a PVC connection between source-destination application pairs are provided later in this document.
  • The Local Management Interface (LMI) message passing is not implemented. The LMI messages, if implemented, help in providing the user with status and configuration information for the PVCs operating over the Frame Relay interface. This feature is not implemented based on the assumption that once the PVC circuits are setup, they are not released until the end of the simulation.
  • The Consolidated Link Layer Management (CLLM) message provides a useful feature to the Frame Relay protocol by acting as an additional congestion notification mechanism. However, since this feature is optional and many systems do not implement this feature, the current model does not support it.
  • The frame check sequence (FCS) checking is not implemented. This feature of the Frame Relay protocol is used to ensure that transmission errors get trapped by the receiving stations. However, assuming that transmission errors are very rare, the implementation of this feature has been omitted from the simulation model.
  • The model provides only a store-and-forward switching methodology (i.e., the frame switching takes place within a network switching device only after it is completely received). There may be instances when the frame switching equipment employs a cut-through switching technology (i.e., the leading edge of the incoming frame may be forwarded on to the next link before the trailing edge is completely received). If cut-through switching needs to be modeled, changes will have to be made to the existing model to achieve the desired characteristics.

Some important features that the model incorporates are:

  • The Frame Relay example model is developed in a modular fashion, with simple interface requirements for DTE and DCE. The FRAD implementation takes care of interfacing the application to the Frame Relay network. Applications may be supported directly on top of the FRAD, or via an interface module if the applications relies on some other protocol. As an example, a Frame Relay-IP interface model (FRIPIF) is provided to illustrate interfacing Frame Relay with IP (an example showing how to set up applications using TCP/IP to support data transfer over a Frame Relay network is also provided).

  • Each PVC supports separate bi-directional specification of all the necessary parameters associated with it (i.e., CIR, Bc and Be along with the source and destination application information). More details on this topic are provided in a later section of this document.

  • The model implements congestion control mechanisms at the DTE on a per-PVC basis by monitoring the data transfer in accordance with the subscription time parameters (CIR, Bc and Be) for that PVC over a measurement interval, Tc. Data in excess of the Bc specification is marked DE, and data in excess of Be specification is discarded immediately.

  • The model allows emulation of zero CIR specification. In this mode all the frames generated at the source are marked Discard Eligible (DE).

  • The model implements congestion notification at the DCE by monitoring network resource utilization in terms of the buffer usage for the outgoing link supporting each PVC. If the buffer usage exceeds a certain preset threshold (to denote onset of congestion), the FECN and/or BECN bits are marked in the outgoing frames. The model allows the user to define the following thresholds (expressed as a percentage of the total bit capacity for that buffer) where congestion control mechanisms are affected:

    • Enable Congestion Status
    • Disable network congestion

    The model also implements congestion control at the DCE by discarding frames which are discard eligible (i.e., with their DE bit set), and in case of severe throttling discarding all the frames until the network recovers from the congestion status. Again, these mechanisms are activated when the buffer usage exceeds predefined thresholds. The following four attributes are used to set these thresholds:

    • Start discarding DE frames
    • Start discarding all frames
    • Stop discarding all frames
    • Stop discarding DE frames

    Additionally, the model provides an option for the network to participate in congestion control by setting the DE bit in frames if the network is in a congested state. This feature can be incorporated in the model by enabling the following parameter:

    • Network action on DE bit

    All of the seven parameters above are attributes of the Frame Relay switching node. The values for these parameters are computed as a percentage of the buffer capacity for that switching node.

  • The model implements Frame Relay fragmentation. The Frame Relay layer fragments application packets that are larger than a specified maximum frame size that the network can carry. At the destination end, these fragments are reassembled to form the original application data packet.

  • The model calculates standard user-information-transfer performance parameters like end-to-end delays, congestion status of the network with respect to time, and the frame residual error rate (RER). More details on these statistics can be found in a subsequent section of this document. Additionally, specific portions of the model can be easily modified to incorporate other custom statistics, if desired.

  • The number of application processes running directly on top of the FRAD can be easily changed without modifying any of the existing models. More details on this topic are provided in a later section of this document.

  • The Frame Relay demand object provides you with a way to define Frame Relay PVCs. You define the PVC as a demand object overlaid between two nodes in a network model, then you can define specific characteristics of the PVC on the Layer 2 Mapping compound attribute of the interface. You do not need to use the Frame Relay PVC Config node in the model.

  • The Frame Relay PVC Config node allows you to define a traffic pattern and assign it to a PVC. You can also create multiple PVCs for a given source-destination pair; this allows you to characterize different PVCs to support specific traffic patterns, thereby providing varying QoS for different higher-layer traffic flows.

  • The standard model library includes a set of Frame Relay "cloud" models that allow you to abstract parts of a network infrastructure. A cloud behaves like a simplified Frame Relay switch with many ports: it supports connectivity between attached devices, but does not include complex operations like buffer management. Clouds allow you to model a subnetwork's packet-loss and latency characteristics while its inner workings remain transparent; see the Standard Model Library chapter of your documentation for more information.