Hewlett-Packard is rolling out a new family of high-performance computing systems that company officials say does not make users have to choose between performance and power efficiency.
The new HP Apollo lineup of Intel-based high-performance computing (HPC) systems will offer up to four times the performance of standard rack servers, but in a smaller and more energy-efficient package, according to John Gromala, senior director of hyperscale product management for HP Servers.
The Apollo 6000 system is air-cooled and uses up to 46 percent less energy than standard blade servers, while the higher-end Apollo 8000 is a supercomputer that comes with a water-cooled design that takes advantage of the superior cooling benefits to liquid without exposing the hardware to any of the risks that come when water gets near electronic equipment, Gromala told eWEEK.
“People want maximum performance, but can’t get maximum performance if the cost is too high,” he said, adding that with the new cooling capabilities of the 8000, those costs are kept lower, opening up what such systems can do and making “HPC available to more people than ever before.”
The new HPC systems are among a range of infrastructure announcements HP officials are making June 9 as the HP Discover 2014 gets underway in Las Vegas. The company also is unveiling new hardware and management offering around its ConvergedSystem solutions.
In addition, the tech giant is making news around software-defined networking (SDN) and storage.
The announcements touch on the broad infrastructure offerings HP is building out for businesses as it looks to grow its IT solutions capabilities and reduce its reliance on PCs.
HP has a significant presence in the HPC space, with as many as 40 percent the systems in the Top500 list of the world’s fastest supercomputers running on HP systems, according to Gromala. However, the top of the list is usually dominated by the likes of IBM and Cray; five of the top 10 systems in the latest list released in November 2013 are from IBM, while another two are from Cray. A cluster system—the Tsubame 2.5 in Japan—that uses a combination of products from HP and NEC is listed at No. 11.
However, Gromala said the new Apollo systems will put the company on par with any other HPC system OEM, both in terms of performance and power efficiency.
“This clearly puts us in play when it comes to [competing] with both vendors,” he said.
Liquids are 1,000 times more efficient at cooling than air, but design issues and the worry about water getting into the electronics have kept adoption of liquid-cooling technologies relatively low, Gromala said. HP’s new system removes the risk and makes the Apollo 8000 the first 100 percent liquid-cooled supercomputer, he said.
The liquid is kept in a contained and sealed cooling loop in the rack, removing the heat that is generated by the systems. The modular rack design of the Apollo 8000 offers up to 144 servers per rack and creates a system with performance that can be four times the teraflops per rack than comparable air-cooled designs, Gromala said. It enables organizations to remove up to 3,800 tons of carbon dioxide waste from their data centers every year, and the heat captured by the cooling system can be used for other tasks, such as heating adjoining offices or rooms.
The Apollo 8000 not only runs on Intel processors, but will support GPU accelerators from Nvidia and Advanced Micro Devices, according to Gromala.
Both the Apollo 6000 and 8000 systems will be available for order starting June 10.
HP Unveils New Apollo HPC Supercomputers
Along with the Apollo systems, HP also introduced its Helion Self-Service HPC, a private cloud solution with a self-service portal that offers a range of resources. The user can manage the solution or have HP manage it for them on a pay-per-use basis. It’s based on HP’s Helion OpenStack cloud platform, which was introduced in May.
In the area of converged infrastructures, HP introduced new tools for managing such environments and reducing operational expenses. The company unveiled enhancements to HP OneView, a converged management platform that enables organizations to use a single tool to handle server, storage and networking elements in tightly integrated solutions.
HP has been aggressively pursuing the converged infrastructure space for several years, competing with the likes of Cisco Systems and its Unified Computing System (UCS). The company in April 2013 announced a ConvergedSystems business unit, and has been steadily building out the portfolio. In December 2013 the company unveiled new systems for virtualization and big data, and in March introduced a system for SAP’s HANA in-memory database technology.
The converged infrastructures are designed to make deployment simpler—organizations aren’t required to integrate the elements themselves—and less costly than traditional environments. HP OneView is designed to better converge the management of the platform, according to Frances Guida, manager of HP’s ConvergedSystems business.
“We’re using software to tie together all of these elements through software,” Guida told eWEEK.
The enhancements to OneView include extending its reach to HP’s 3Par storage lineup and supporting flexible pools of storage through automated converges of servers and storage. In addition, there is new integration with management systems from VMware and Microsoft.
HP also is combining OneView with HP’s ConvergedSystem 300 and 700 solutions for virtualization, which will offer twice the performance of competing systems at lower cost and complexity, according to company officials.