If middleware was the theme from last week, then this week is all about hardware, from networking/interconnects (Arista) to systems (HP) to storage (Panasas). Let’s look at little closer at this week’s top stories, starting with HP.
HP’s rebooting of the Apollo brand this week was well received. This new family of high performance computing systems aims to balance high density with energy efficiency, and raw performance. Called “HP Apollo” – although HP says there’s no connection to the workstation firm that it acquired back in 1989 – the line includes two new “Apollo” server systems: the air-cooled Apollo 6000, intended for dense, enterprise deployments, and the liquid-cooled Apollo 8000 for large-scale high-performance computing.
After the Apollo launch, HP still had another announcement up its sleeve. At the Discover 2014 conference in Las Vegas, HP CEO Meg Whitman and Martin Fink, its CTO and the head of HP Labs, unveiled a massive new architecture project, referred to as the Machine.
“Why do we call it The Machine?,” queries Fink. “When we first started developing it, we wanted to be very careful not to call it a server, workstation, PC, device or phone, because it actually encompasses all of those things. So as we were waiting for Marketing to come up with a cool code name for the project, we started calling it The Machine—and the name stuck.”
So what is the Machine? In a nutshell, it’s a new computer architecture based on memristors and silicon photonics. In his feature coverage at Enterprise Tech, Timothy Prickett Morgan elaborates:
“With the Machine, HP is going to tightly couple custom-tuned processors with its memristor non-volatile memory, which has been in development for the past six years, to create a system that eliminates the nine to eleven stages of memory and storage hierarchy that are in a server these days.”
Fink said the concept can be summarized in six words: electrons compute, photons communicate, ions store.
The project has the attention of 75 percent of HP’s R&D division, HP Labs, and HP’s roadmap shows the Machine coming to market over the next two years.
On Tuesday, Panasas, a vendor of scale-out NAS, announced enhancements to its ActiveStor 16 hybrid storage appliance and to its PanFS 6.0 parallel file system. The PanFS 6.0 file system now features RAID 6+ triple parity data protection, and the ActiveStor 16 array employs 6TB HGST UltraStar He6 helium-filled drives and a larger 240GB solid state disk.
“With big data workloads pulling high performance computing into the mainstream of the enterprise, customers’ expectations for reliability and availability have increased at a time when traditional data protection has reached its limits,” said Steve Conway, research vice president, IDC. “The developments in Panasas ActiveStor 16 and PanFS 6.0 show timely industry leadership in addressing this growing concern and tackling the challenges of reliability and availability at scale.”
Last Friday, Arista Networks, Inc. announced the pricing of its initial public offering of 5.2 million shares of its common stock at a price to the public of $43.00 per share. The company is valued at around $2.75 billion and the IPO generated about $226 million. Currently, the stock is trading at $66.
In covering the story, Editor Nicole Hemsoth writes “To be clear, other than having a heavier warchest to throw at new Ethernet efforts, Arista’s role in HPC will remain small, although for some workloads in oil and gas and financial services (especially where InfiniBand has been foregone in the past due to a lack of latency and bandwidth sensitivity) it will remain a top choice.”
Notable HPC use cases for Arista technology include BP’s 2.2 petaflop system and the San Diego Supercomputer Center’s 5.5 petabyte-capacity cloud storage system. Arista has a similar deployment at Los Alamos National Lab through a partnership with Panasas. “The real stories about Arista’s success in HPC datacenters, however,” writes Hemsoth, “is locked behind the guarded walls at high frequency trading shops, among others, where the network’s secret sauce is part of a much larger approach to ultra-fast computing that we’ll never hear about.”
A recent report from Intersect360 Research shows the worldwide HPC market expanding at a 4.6% compound annual growth rate over the next four years in spite of mixed signals from vendors and lethargic public sector spending. Intersect360 Research’s newly released Worldwide High Performance Computing 2013 Total Market Model and 2014–18 Forecast states: “despite weakness in the public sector caused by government austerity programs, commercial vertical markets were strong enough to drive 2013 HPC revenue to $30.5 billion worldwide, 2.5% growth over 2012.” With continued recovery putting wind in its sales, the HPC market should reach $38.1 billion by 2018, according to the research firm.
Other interesting data points from the study:
- Cloud computing market is forecasted to grow 18.6 percent during the same time period (2014-2018)
- For the first time, cloud comprises a full percentage point of the HPC market.
At the national research level, a working group of the National Institutes of Health is requesting funding for the US Brain Initiative, to the tune of $4.5 billion dollars. Champions of the program maintain that the amount is inline with other grand challenge programs. For example the Human Genome Project, the 10-year project that resulted in the first human genome being sequenced in 2003, had a price tag of $3 billion. The NIH is expected to award its first BRAIN grants this September. There will also be opportunities for US BRAIN project researchers to coordinate efforts with its European counterpart, the Human Brain Project, which has been awarded $1.3 billion over 10 years.
On the Road
International Supercomputing Community to Gather in Leipzig From June 22-26