As high-performance computing (HPC) moves beyond the science labs and oil rigs and into mainstream enterprise IT, technology service provider HP has beefed up its HPC portfolio and is targeting supercomputer builders. But is it leaving its rivals behind in the HPC space?
HPC involves the use of parallel processing to run advanced application programs efficiently, reliably and quickly. It usually refers to systems that function above a teraflop or 1012 floating-point operations per second.
According to a report by analyst firm IDC, HP led the $10.3bn HPC industry in 2013 with a 35% share of the HPC server market. IBM was the second-largest provider with a 23.1% market share and Dell was third with 17.2% . Other major players, such as NEC and Sugon, also made strong year-on-year revenue increases, driven by their gains in the supercomputer segment.
Earl Joseph, program vice-president for technical computing at IDC, said: “HPC technical server revenues are expected to grow at a healthy rate because of the crucial role they play in economic competitiveness as well as scientific progress. As the global race toward exascale computing fuels the high end of the market, more small and medium-sized businesses and research organisations are exploiting HPC servers for advanced simulations and high-performance data analysis.”
IDC predicts that the supercomputer segment will see modest growth for the rest of 2014, but will expand at a compound annual growth rate of 7.2% through to 2018.
A supercomputer is a machine that performs at, or near, the currently highest operational rate for computers. It is typically used for scientific and engineering applications that must handle very large databases or do a vast amount of computation, or both.
Currently, China’s National University of Defense Technology’s supercomputer Tianhe-2 – powered by Intel – is rated the world’s number one system (for the third consecutive year) with a performance of 33.86 petaflops – or quadrillions of calculations per second.
HPC is moving out of labs, oil and gas rigs and government agencies to enterprise IT
Philippe Trautmann, EMEA sales director, HPC and datacentres, HP
Philippe Trautmann, HP's EMEA sales director, HPC and datacentres, said at the International Supercomputing Conference (ISC) in Leipzig, Germany: “We are seeing that HPC and supercomputing are the fastest-growing market segment in enterprise IT and that is why they are becoming a core focus for us.”
HP’s own research found that hyperscale computing is growing by more than 20% in value every year. “HPC is moving out of labs, oil and gas rigs and government agencies to enterprise IT,” said Trautmann.
At the ISC and at HP Discover Las Vegas, HP showcased its Apollo server systems, which provide rack-scale performance and energy savings as well as HPC-as-a-service, aimed at the HPC market.
The first product in the range, the Apollo 6000, is an air-cooled system that can pack 160 Intel Xeon servers per rack while using almost 50% less energy than previous types of server. HP is targeting the server at public cloud providers.
The second version, the Apollo 8000, offers performance peaks of over 250 teraflops per rack and can be configured with 144 servers per rack. It is aimed at the high-end HPC market, including large government agencies and universities. John Gromala, director, modular systems at HP, said: “It offers up to four times the teraflops per square foot and up to 40% more FLOPs per watt than comparable air-cooled servers.”
The Apollo 8000 is an entirely water-cooled server system that offers further energy efficiency. It uses a technology called “dry disconnect”, which means it does not let water touch the machine but uses sealed heat pipes to allow treated water to run between the cores, draining off heat more efficiently than an air-cooled system. Scott Misage, director for HPC at HP, said: “Water cooling may be a common approach, but we are the only ones currently using a warm water cooling approach.”
Gromala added: “With our water-cooling system, you don’t have to cool the water first to use it for IT cooling. That itself saves enterprises huge power costs. It also helps reduce your carbon footprint.”
HP’s timing is perfect and could help it gain significant market share
Steve Conway, research vice-president, IDC
The Apollo systems will become generally available from January 2015 and HP is looking to boost its fortunes through HPC at a time when hardware and server business continue to dwindle.
Steve Conway, IDC’s research vice-president, told Computer Weekly at the ISC: “Until now, HP did not even have products to compete in the high-performance computing space – and HP itself admits that.” Also at the ISC, Misage said: “Yes, until now [before the Apollo range] we did not have anything high-end for HPC.”
But today, HP's offerings can compete with IBM and Cray, two of the biggest players in the HPC space, said Conway.
The timing of HP’s launch of HPC products is crucial in helping it gain market share in the segment, said Conway.
“HP’s timing is perfect and could help it gain significant market share,” he added. “HP’s biggest competitors in this field are IBM and Cray.”
HP’s Apollo systems compete with IBM’s large water-cooled systems or Cisco’s UCS, said Conway, “but IBM is still finalising the sale of Lenovo and is busy with that, so HP can gain some ground”.
Another analyst, Addison Snell from Intersect 360, said: “HP and Dell are grabbing market share from IBM in the HPC space. But who will grow more will depend on who gets more from IBM in the next few months.”
Snell added: “But HP is focusing on the right segment at the right time. HPC is registering very strong and stable growth in an otherwise tricky enterprise IT marketplace. And there isn’t one HPC project that isn’t cool – it is used by engineers and scientists to do things that matters to us, from shinier hair to safer cars.”
According to Snell, 44% of current HPC users are planning to increase the scale of their projects.
There isn’t one HPC project that isn’t cool – it is used by engineers and scientists to do things that matters to us, from shinier hair to safer cars
Addison Snell, analyst, Intersect 360
HP is aiming to boost volume sales with its Apollo 6000 range and provide HPC-as-a-service on the cloud to small and medium businesses that cannot afford on-site supercomputers.
Two organisations, the University of Tromso in Norway and the US Department of Energy’s National Renewable Energy Laboratory (NREL), use the Apollo 8000 system. “We wanted to do HPC inside our own datacentre and energy efficiency was crucial for us,” said Steve Hammond, director of NREL's Computation Science Center.
“Ours is one of the most energy-efficient datacentres in the world and we don’t have any mechanical compressors for cooling water. So the warm water cooling offered by HP Apollo was ideal for us.”
NREL’s current datacentre HPC contract with HP is worth $10m and it is already planning to double the size of its datacentre estate using Apollo servers. “We currently run 1,400 nodes and 1.2 petaflops of data in our HPC environment,” said Hammond. “By the end of this year, we will add another petaflop and 1,200 nodes.
“Our investment in HP is already yielding us return on investment in terms of cooling efficiency.”
The Apollo system is also enabling NREL and the University of Tromso to save up to 3,800 tons of CO2 emissions per year.
The University of Tromso is able to achieve even more savings by using the warm water from the Apollo systems to heat its rooms while temperatures outside are sub-zero.
Meanwhile, Intel has placed an order for hundreds of Apollo systems from HP. Analysts predict that the next edition of the TOP500 list of supercomputers in the world at ISC could feature some machines powered by HP’s systems.
Paul Teich, CTO and senior analyst at Moor Insights & Strategy, writes on the company blog: “HP’s HPC technologies, including the Apollo systems and The Machine – the computer architecture based on memristors and silicon photonics – are about as disruptive as new tech can get.”