ATDAPS is a powerful, PC-based, modeling and simulation tool suite for design and performance evaluation of distributed and parallel computer and network architectures. It simulates real time network traffic at the packet level and computes key performance metrics such as latency, throughput and resource utilization. ATDAPS allows an analyst to rapidly evaluate alternative architecture designs such as

• Distributed vs. Centralized Processing
• Distributed vs. Centralized Data
• Number and Capacity of Network Resources

Click on image to enlarge (47k)

The ATDAPS Suite includes three prototype tools that can be run in sequence

1. A Partitioning Tool allows a user to graphically define a distributed/parallel communications computing system configuration and its environment based on architecture related system requirements (e.g., minimize processing time, minimize number of processors). Partitioning involves defining specific sets of processes to be performed by individual distributed and/or parallel processors (nodes).
2. A Mapping Tool then allows a user to graphically map the executable application software to the partitioned processors using performance-based criteria (e.g., speed of computation, processor utilization, number of processors used, precedence of execution).
3. An Assessment Tool (VHDL* Simulator) simulates execution of the partitioned and mapped distributed/parallel application software, as hosted in the user defined environment. It then predicts system performance in terms of key parameters (e.g., latency, throughput, resource utilization).

ATDAPS allows a user to iteratively refine partitioning and mapping solutions by running the above three tools multiple times in sequence. This allows the user to improve the design based on iterative feedback from the Assessment Tool's performance predictions.

* VHDL = VHSIC Hardware Description Language

ATDAPS is intended to be integrated with commercially available CASE tools to allow users to quickly and accurately design, analyze and assess candidate distributed/parallel computing system architectures and to define/refine a system design to create an efficient system solution for a particular processing problem.

CACI MTL used ATDAPS to assess the impact of adding a network monitoring function to a commercial company's large scale campus network. ATDAPS predicted network performance under various combinations of data management schemes with both centralized and distributed computing architectures. It then identified the optimal solution, which achieved the added functionality with no degradation to network services.

In another application ADTAPS predicted an optimal parallel computing configuration for a real time, electronic warfare simulation of hundreds of simulated combat systems and their electromagnetic interactions.

In both cases ATDAPS identified optimal system configurations rapidly and easily before any new hardware was purchased, thus saving the customers significant time and expense.

Sponsor - Air Force Research Lab - Dayton, OH

Download brochure ppt - 255 KB