Understanding IP QoS Model Behavior
Both DES and flow analysis support IP QoS, however, due to the nature of these technologies, there are differences to the way QoS behavior is modeled by each type of simulation. Understanding the way each model behaves will help you interpret your simulation results and enable you to decide which technology best suits your modeling objectives.
This section highlights the behavior of flow analysis and DES for the following aspects of the IP QoS model:
Traffic Representation
The level of detail to which you can characterize the traffic in a network accounts for much of the differences between the way DES and flow analysis model IP QoS.
In DES, each packet is modeled individually and the size and other characteristics of the packets vary according to the settings configured in the application that created the packet. For example, if you use distributions to vary the size and inter-arrival time of the application’s traffic, its packets will be of different sizes and will be sent out at different intervals.
In flow analysis, packets are not modeled distinctly. Instead, traffic flows are used to summarize some of the traffic between two nodes. All of the traffic in a flow is configured to use a single average packet size and a single average inter-arrival time.
Because of the amount of detail that flow analysis has about the traffic it is modeling, it does not account for differences in the traffic in the same way that DES can. DES, for example, can account for the performance variations that occur when several large packets arrive in a short amount of time. Flow Analysis, on the other hand, will not capture that behavior since it operates using average values only.
Queuing
Both flow analysis and DES model queuing behavior, but they do it in different ways.
Flow analysis does not model routing or topology changes during a simulation—it assumes the network is in a steady state. Because of this, it can use mathematical formulas (M/G/.) to model queuing behavior. These formulas assume an exponential inter-arrival time and constant processing time.
Discrete event simulation lets you model failure and recoveries in a network and computes the routing changes that result. The behavior of each packet is modeled with the current state of the network. Queuing behavior is also modeled in detail: each queue is modeled using the configured queuing scheme and all of the packets going through the queue are modeled distinctly.
Policing
Policing is modeled in DES, but not in flow analysis.
Shaping
Traffic shaping is modeled in DES, but not in flow analysis. The following types of traffic shaping are supported.
- Generic Traffic Shaping, which uses flow-based weighted fair queueing. Access lists are used to determine which traffic to shape.
- Class-Based Weighted Fair Queueing (CBWFQ), which is traffic shaping configured per class rather than within an access list, is supported in both of the following ways
- In conjunction with GTS
- Inside GTS
Note: Distributed traffic shaping is not supported.
Congestion Avoidance
Congestion avoidance—RED and WRED—are modeled in DES, but not in flow analysis.
Reporting
The type of information that you can obtain from a simulation also varies according to the type of simulation. The main difference is that only DES collects response time statistics.
The following statistics are available for analyzing IP QoS behavior. The statistics are available for both types of simulation, unless otherwise indicated.
- response times (DES only)
- packet end-to-end (ETE) delay
- packet ETE jitter (both, but flow analysis gives only the maximum)
- packet ETE loss, maximum (flow analysis only)
- queuing delay
- packet loss
- jitter