As virtualized computing environments become more widely accepted across the public and private sectors, those in the scientific arena have begun to take note of the technology's potential. With data center optimization efforts continuing to consolidate servers and streamline the configuration of tech assets for increased efficiency, researchers may now be able to achieve even greater performance with a software-defined network approach. Virtualization is no longer reserved for the business world or federal databases - its power could bring scientists closer to the next big discovery.
Hopkins engineers stuck in second gear
Perhaps the most demanding, dynamic hardware requirements are found in the world of simulation laboratories. High-performance computing is a notoriously complex and challenging field, and engineers are expected to push their systems to their absolute limits to facilitate simulations and other experimental initiatives. Such projects have long been orchestrated with the use of massive hardware units and unwieldy networking demands, but according to a recent article from GCN, a shift toward virtualized infrastructure has simplified this structure and improved performance benchmarks.
At the Johns Hopkins University Applied Physics Laboratory Air and Missile Defense Department's Combat Systems Development Facility, legacy systems were preventing engineers from tapping into the full potential of their resources. The lab's systems are constantly under the strain of highly demanding simulation tasks, and according to Edmond DeMattia, a senior system engineer and virtualization architect at the facility, its current configuration was simply not making the grade when it came to the dynamic requirements of the organization's projects.
"We had two stovepipe systems with one running Windows and one running Linux," DeMattia told the news source. "There are 1,500 cores per cluster, and everyone was sharing that computer using grid scheduling. Some simulations take five seconds per task, and we run that same task up to a million times. While others may take 15 hours per task, but are only run 1,000 times."
Virtualized setups save the day
Facing major inefficiencies and reeling from the reality of wasted resources, DeMattia and his team employed ESXi, a VMware hypervisor which manages virtual machines from a centralized point of control. By adding this virtual layer, the lab was able to access the computing power it had been missing by relying on traditional configurations. The tool allowed various operating systems to utilize unused nodes from other elements of the infrastructure, achieving greater performance and maximizing the lab's computing efficiency.
"My team fundamentally redesigned how high-performance scientific computing is performed in the Air and Missile Defense Department by utilizing virtualization and distributed storage as the framework for pooling resources across multiple departments," DeMattia said, according to GCN.
Not only was the facility able to enjoy faster simulations by taking a virtualized approach, but it also reaped cost benefits from the heightened efficiency of its systems, the source noted. Since new computing nodes can cost up to $10,000 each, not to mention an additional $40,000 for cooling and energy, squeezing more power out of its existing resources was a major power play in terms of savings. GCN estimated that the facility has been able to save around $504,000 simply by adding a virtual layer to its setup.
Yielding results in the field
What does the Combat Systems Development Facility's improved simulation performance mean for the nation's defense technology? The Albuquerque Journal highlighted a recent advancement in IED detection aided by the use of unmanned aerial vehicles. The system is able to deliver high-resolution imagery revealing irregularities in the earth where explosive devices may have been buried. While the Hopkins lab oversaw only a part of the technology's realization, the organization's contribution will prove invaluable to the protection of troops and civilians overseas.
.png)
