Dell has upgraded its line of PowerEdge servers aimed at high-performance computing applications and revealed 27 racks of the new HPC servers make up the Comet supercomputer launched recently at University of California, San Diego.
Based on Intel Xeon E5-2600 v3 processors that provide up to 18 cores per socket to make available 144 cores per 2U chassis, the new PowerEdge C6320 servers can be configured with up to 512GB of DDR4 memory and up to 72TB of local storage.
Brian Payne, executive director for Dell Server Solutions, said that in addition to being used to drive traditional HPC applications, the servers are also being applied to run Big Data analytics based on Hadoop and configured as hyper-converged appliances designed to run VMware EVO: Rail software or hyper converged Nutanix software that Dell also resells.
As Intel continues to move up the performance curve of its processors, IT organizations are increasingly opting to run application workloads that were once the sole province of RISC servers running Unix or custom-built supercomputer platforms on Intel Xeon-class processors. According to Payne, that shift not only expands the number of HPC-class applications that IT organizations can afford to run, it expands the market place for Dell into new segments that the vendor previously would not have been able to play in.
“We’re not only seeing customers deploy these servers in hyper converged appliances,” Payne said. “Customers are also running Hadoop on them to save money by offloading tasks from mainframes.”
The PowerEdge C6320 comes integrated with iDRAC8 with Lifecycle Controller, which enables IT organizations to automate routine management tasks to reduce the number of steps required to deploy, monitor, and update their servers.
All told, Dell claims that the new HPC servers provide up to two times performance improvement over the previous generation of servers running a leading HPC benchmark and can achieve 999 Gigaflops on a single server.
In the case of UCSD, the school has deployed a total of 1,944 nodes or 46,656 cores, which represents a five-fold increase in compute capacity over the previous HPC system the university was employing.
The National Science Foundation gave the university a $12 million grant in 2013 to build the Comet system.