dcsimg

Getting to the Core: HPC Cluster Design & Optimization

All clusters are not created equal. Building an efficient HPC cluster requires a careful match between hardware resources and application needs. Cluster designs based on core count or clock speed are likely to underperform, cost too much, or both.

All clusters are not created equal. Building an efficient HPC cluster requires a careful match between hardware resources and application needs. Cluster designs based on core count or clock speed are likely to underperform, cost too much, or both.

In this webinar, Hewlett Packard and AMD will present four considerations for HPC cluster design that are critical for maximizing core efficiency and scalability:

1. Application characterization – How to measure performance and identify resource bottlenecks

2. Processor selection – Mapping application requirements to processor options for best price/performance

3. Memory configuration – Designing the memory subsystem for optimal performance

4. System design considerations for peak performance – Networking, scalable management, and energy efficiency

After specifying the best cluster design, what are some performance optimization techniques to improve data locality and tune your cluster for optimal results? HP and AMD will provide valuable insight on this topic and will also share the newest features in the AMD Opteronâ„¢ processor architecture and HP Cluster Platform offerings that can assist in other areas of cluster optimization.

Comments are closed.