Verify Hardware Architecture in a Network

Click here to view and execute the VisualSim model

The biggest challenge in product engineering is to validate the implementation in the operating environment. Products are always used in ways that were never intended. It is essential to gain a detailed understanding of the environmental impact be it on performance, functionality or at the user-level. In this model the dual processor server is introduced as the hardware in Node 1 of the Route Optimization System. This is a key benefit of VisualSim as the protocols are combined with the implementation architecture.

Project Objectives
The dual processor server was designed to operate as a server blade for networking applications. The blade was designed in solitude. To get an accurate evaluation of the utilization, the design had to be tested in a network environment. The generic implementation of Node 1 was replaced with the detailed implementation from the Dual Processor System. The workload was maintained the same.

Capabilities Demonstrated
This model shows the use of the Node blocks for non-infrastructure purposes. The Node can be used to connect architectures to a network topology. This model also demonstrates mixing different layers of abstraction in a single model to conduct system coverage. The model shows the flexibility of the SmartBlocks to be connected easily into a new system with limited modifications and is another innovation of the SmartBlocks. The networking SmartBlocks can be used to define very detailed network architectures without requiring extensive coding.

 

Model Statistics
Number of unique blocks required to create the model = 10 + 1 Hierarchical block (Dual Processor Server)
Time to do the initial model construction = 5 Hours
Model analysis and refinement = 2 Days
Documentation = .5 day

Model Construction
This model is identical to the Route Optimization System. The only difference is that the Dual Processor Server example from the system architecture section is introduced as the implementation of Node 1. The Workload is the same as in the Route Optimization example. The parameters for the implementation of Node 1 are the same as for the Dual Processor Server.

Results
There are four sets of results published with this model-

The results here must be compared with those of the Dual processor Example. The processor utilization is about 1.2% here as opposed to 23% when simulated in solitude, for the same traffic profile. This shows that the network analysis is key to understand the impact of the protocol overhead and the performance constraints created by the environment. All the devices connected to the server blade cannot execute at the same speed as the server. Hence, even though, the Dual Processor Server, can perform at 10 Gbps, there is no network environment that can feed data at that wire-rate. The architecture can be cost-reduced by using slower parts and still maintain wire-speed.

When the workload is increased by reducing the Task_Rate, you will notice that utilization increases at a much slower rate than the stand-alone analysis. The model can be further expanded to study the additional overhead caused by the protocol stack, retransmissions etc.