In mission critical projects, our client achieved manifold improvement in speed by harnessing an energy-efficient water-cooled High Performance Computing (HPC) solution - to solve their complex computational research experiments.
"We've worked with StarOne IT for 4 years… and frankly the HPC solution they have implemented for us gives us the much required computing power and bandwidth performance that we required for our mission critical project. We are very happy with the professionalism shown by StarOne IT" - Director, Client Organization.
Overview
Large enterprises like space research organizations have mission critical systems to solve very complex computational problems. It is through data that groundbreaking scientific discoveries are made, game-changing innovations are fueled, and quality of life is improved for billions of people around the globe. Legacy applications will have to look for opportunities to modernize IT systems to enable parallel computing to improve computing power. High-performance computing (HPC) is the ability to process data and perform complex calculations at high speeds, to fuel scientific, industrial, and societal advancements.
Challenges
Our client, which specializes in space exploration wanted to install a state-of-the-art computing solution to handle its ever-growing compute demand for research and analysis. Their compute intensive analysis software used complex logic written in Ansys, CFD++, and FORTRAN. Research cases that used to be run in labs required very high throughput and very low latency. These cases ran on high-end workstations, via InfiniBand. However it had constraints both in terms of performance and IT infrastructure to run the workloads. Each Research cases had to be manually submitted and Scientists had to wait till completion, to run the next case. This cases would require weeks to complete, thereby prolonging time for research & analysis, as well as decision making.
Solution Highlights
Following multiple technical evaluation of various HPC solutions available, client decided to partner with StarOne IT to implement a modern day energy-efficient HPC solution to help power their for research operations. StarOne IT deployed a high-speed, scalable Lenovo based Water Cooled HPC Solution with a computational capacity of 80 TFLOPS.
Computational Capacity : 80 TFLOPS peak
No of Nodes : 72 Nodes
HPC Solution: Lenovo Direct Water Cooled HPC system
Cooling Capacity: Water Cooling Capacity - 30KW, Air Cooling Capacity - 14KW
Energy requirements: UPS Capacity of 70KW (40KW for IT + 30KW for Non IT)
Benefits
With the addition of extra 80 TFLOPS, Research & Analysis got off to a flying start. They were now able to complete compute-intensive data analysis & experiments much easier, in lesser time. With the new HPC cluster deployed by StarOne IT, Research Cases, which once used to take weeks to complete, would now take just a couple of hours to finish. They had the ability to scale-up compute resources to upto 80 TFLOPS almost immediately. Also multiple jobs could be scheduled to run 24x7, without need for any human intervention or issues of over-heating of Servers. Users just had to submit the cases for execution; based on job priority and resource availability the scheduler would run the jobs automatically. All this helped the client conduct extensive component experiments for critical space mission, which was able to be completed in record time, with the help of the newly installed HPC Servers.
The superior heat conducting efficiency of water- cooled Compute Nodes, help critical server components operate at lower temperatures compared to air-cooled systems. The WCT compute nodes operate at normal atmospheric temperature i.e. ~33-35 degree Celsius and can work upto 50 degree Celsius without performance degrading, thereby reducing the operational cost. With up to 90% heat removal efficiency, the HPC provides up to a 40% savings in data center energy expense. Also with a lower footprint/space to operate the high density compute nodes, it meant the HPC water cooled Server Racks took lesser space when compared to the space occupied by the bulky workstations.
StarOne IT's professional support services also ensured that client did not have any unplanned downtime. Client fully relied on StarOne IT's expertise in HPC to help them implement a system for their critical missions. For the organization’s success, StarOne IT received high commendations from the client, and we continue to help them implement & support more HPC installations at their various data centers.