The fact of the matter is, to be blunt, that building, maintaining, and optimizing systems for high-performance computing (HPC) is different than any other part of the IT world. That is why most companies engaged in simulation who use HPC often struggle with their computers and networks. There is nothing wrong with their IT departments, they simply don’t have the manpower or the experience to support HPC systems. And that is why PADT offers IT support services tailored towards the needs of simulation users. We bridge that gap between the unique needs of HPC for simulation and customer’s existing IT infrastructure.
We started offering this service simply because customers asked us to. As part of ANSYS technical support duties, we kept getting calls from customers who were just not getting good performance out of some very expensive computer hardware. After looking into it we often found that they had the wrong hardware, it was configured wrong for HPC, or there was unnecessary overhead on the system. In each case, we got together with the customer’s IT department and the users to understand the problem and implement fixes.
PADT’s IT team can offer a variety of services, including, but not limited to:
Focused on Performance
Most computer systems and the IT infrastructure that supports them, in the commercial world, are focused on security first, operating cost second, then performance. And these are the right priorities. But in the HPC world, those same systems and infrastructure have to be focused on performance, with robustness second. Security is important, but you solve that problem by isolating the systems and controlling access. PADT’s IT team gets this and because we run a large HPC infrastructure for our own simulation consulting business, they know how to set things up right and keep them tuned for performance.
A Partnership Between Users, IT, and PADT
Our deep knowledge of scientific computing and the hardware it runs on is only one part of our success in this area. The other is knowing how to be that translator between IT and the users of demanding numerical software packages. We understand why IT can or can not open certain ports, and why the user needs those ports. We get the desire to establish a company-wide policy using remote drives that use RAID 1. We also know that doing that kills HPC performance. That starts the conversation on why users need local drives in RAID 0 for their number crunching.
The examples go on and on and involve memory, network fabric, different versions of MPI for each solver, and much more. Often times just being able to explain these issues in terms everyone can understand is the greater value PADT can add. And then our IT experts will be there with the customer’s IT experts at 11:30 PM installing those GPU cards and configuring them.
Maximize the Return on your HPC Investment
The cold hard reality is that companies spend large sums on advanced hardware, software, and infrastructure. Why not spend a little more to make sure you are getting the most from that sizeable investment. PADT is here to help, to work with your IT and users, to get the most out of the tools you have. Please don’t hesitate to call us at 480.813.4884 or send an email to email@example.com to learn more.