July 09, 2013
The relationship between green computing and high performance computing is an evolving one. For example, while no supercomputer occupied top ten slots in both the Top500 and Green500 lists (like Titan did in the previous iteration), five machines occupied top twenty slots in both lists. The focus remains on either performance or efficiency, with an overlap starting to take hold.
Cloud computing might well serve as a bridge between the two, as top HPC management layers couple with known efficient datacenters, such as what happened when Datapipe and Verne Global formed a partnership in the middle of June that will bring the Stratosphere management platform to Verne’s Icelandic datacenter.
“Having the Stratosphere environment here within our datacenter allows us to sell not only to the customers for the space that they might need,” said CTO of Verne Global Tate Cantrell to HPC in the Cloud, “but also ensure that they can engineer their applications and their solutions such that they can really burst in to this environment and take care of the spike without leaving the four walls of the facility.”
Stratosphere itself is API-driven, and utilizes 100 percent Solid State Drive storage. The platform enables a maximum of 32 physical core equivalents per instance along with a half terabyte of RAM and tens of thousands of IOPS per volume in a 10GE network. It is most utilized as an elastic cloud environment in which to implement Apache CloudStack.
“Iceland is a great place to run high performance applications,” said Ed Laczynski, Senior Vice President, Cloud Strategy & Architecture at Datapipe, to HPC in the Cloud. “With the energy security, you know you’re going to be running applications that consume 90 percent plus of your CPUs on a hypervisor.”
According to Laczynski, the types of applications Datapipe plans to host on the Verne Global servers will involve a significant amount of big data analytics. “We’re running a lot of intensive applications, big data applications so a lot of Hadoop, a lot of MongoDB, a lot of Vertica and analytics type applications.”
Among those applications is that of an unnamed financial services client of Datapipe’s. Laczynski expanded on the goal of their application that is already set to run in the Icelandic datacenter, noting that “they’re putting terabytes of data from their affiliates and customers around the world into the datacenter, processing that data, reducing it, and adding their business value to that data, and submitting it back out to their partners and customers.”
Introducing Stratosphere into the Icelandic datacenter was relatively simple. Laczynski noted that there were no major challenges that Datapipe had not faced in previous implementations in AWS or in their own datacenters.
The latency between Iceland and London, and to a lesser extent between Iceland and the United States, is a feature, according to Laczynski, that makes Iceland attractive for setting up a cloud infrastructure that is energy efficient and able to run high-performance applications.
“On a network front," Laczynski said, "it’s a great place for our customers that are doing business in Europe and the US. It’s a great business continuity site between the two continents. There’s really good latency numbers from Iceland to London.”
The energy efficiency advantages and the latency prospects combined are among what customers are looking for in choosing the Icelandic path over options like Amazon Web Services or Microsoft Azure. That latency makes it easier to pull data out of Iceland, a necessity for institutions looking to run fast-paced HPC applications in the facility.
Additionally, according to Laczynski, going with the Datapipe package at Verne versus similar platforms at an AWS datacenter in Ireland offers stability and business continuity for long term storage. “A lot of what we’ve brought to the table with Verne is the ability to give customers security, energy, and capacity.”
Of course, with the energy concerns presented through the push to exascale and the rising prominence of the Green500 list, there are worse ideas than setting one’s cloud-based HPC in the Icelandic datacenter.
Verne Global is well-known over on Green Computing Report, as their Icelandic datacenter has received praise from the green computing community for reportedly running on 100 percent renewable energy. Further, according to Verne’s Tate Cantrell, the facility utilizes air and water cooling techniques that eschew the necessity for additional energy-consuming machinery. As Cantrell noted, that reduces facility costs for both Verne and their customers by extension.
“We’re able to provide 100 percent air-cooled,” Cantrell said, “or potentially water cooled solutions that don’t rely on additional equipment to provide that cooling so this ends up reducing costs not only for the energy but also for the facilities charges themselves, which allows us to be competitive with other cloud markets.”
Green is a color that is starting to look more and more attractive on high performance machines. Cloud services, managed by Datapipe’s Stratosphere layer, from energy efficient datacenters like Verne’s Iceland could play an important role in the evolution and accessibility of green HPC.
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?