Abstract

The use of a network of shared, heterogeneous workstations each harboring a reconfigurable computing (RC) system offers high performance users an inexpensive platform for a wide range of computationally demanding problems. However, effectively using the full potential of these systems can be challenging without the knowledge of the system's performance characteristics. While some performance models exist for shared, heterogeneous workstations, none thus far account for the addition of RC systems. Our analytic performance model includes the effects of the reconfigurable device, application load imbalance, background user load, basic message passing communication, and processor heterogeneity. The methodology proves to be accurate in characterizing these effects for applications running on shared, homogeneous, and heterogeneous HPRC resources. The model error in all cases was found to be less than 5% for application runtimes greater than 30 s, and less than 15% for runtimes less than 30 s.

Keywords: Analytic performance modeling; Reconfigurable computing (RC); High performance computing (HPC); Performance evaluation

Article Outline

1. Introduction

1.1. What is HPC, RC, and HPRC?

2. Model development

2.1. HPRC multi-node analysis

2.2. Load imbalance model

2.2.1. General load imbalance model

2.2.2. Application load imbalance model

2.2.3. Background load imbalance model

2.2.4. Complete load imbalance model

3. Model validation

3.1. Validation applications

3.1.1. SAT solver

3.1.2. Matrix vector multiplication

3.1.3. AES algorithm

3.2. Validation measurements

3.2.1. Homogeneous resources

3.2.1.1. SAT solver

3.2.1.2. Matrix vector multiplication

3.2.2. Heterogeneous resources

3.2.2.1. SAT solver

3.2.2.2. Matrix vector multiplication

3.2.2.3. AES algorithm

4. Model applications

4.1. Minimizing impact to other users

5. Conclusions

References

Vitae

This research was partially supported by the National Science Foundation (grant number CCR-0311500) and the Air Force Research Laboratory (contract number 601-01-S-0086).