This page contains a Flash digital edition of a book.
HPC 2013-14 | Te cloud The silver lining?


Beth Harlen looks to the cloud for the future of HPC


Te definition of what constitutes ‘cloud’ will oſten depend on whom you ask. Te term cloud computing covers a considerable range of services that are being delivered under that banner. Tere are several service models to choose from, such as infrastructure-as-a- service (IaaS), soſtware-as-a-service (SaaS), and platform-as-a-service (PaaS) (see box ‘Selecting a service’), as well as a range of deployment models that include public, private, community, and hybrid clouds – each with its own attributes. Dave Kress, CEO at Coraid, offers a


good definition of the differences between the ownership models: public clouds offer outsourced services that allow businesses to deploy resources quickly and easily when needed. For IT customers who want the agility and ease of use that public clouds provide, but with added security and control, private clouds provide an attractive alternative. Finally, hybrid clouds promise a flexible and fluid middle ground between private and public clouds. Kress believes that public cloud service


providers such as Amazon have set the standard for agility, ease of use, and price for infrastructure, and that other service providers are now tackling the challenge of providing differentiated services at competitive prices while keeping costs low. ‘One approach for differentiation is to offer tiers of services with different performance and availability tiers at a variety of price points. Another is to offer value-added services beyond basic infrastructure,’ he said. Adding to the mix – and causing a level


of confusion in the industry – is the distinction between generic, non- HPC clouds and those offered by HPC providers. While the former offer a cost-


14


effective solution, they lack the specialised infrastructure provided by the latter – an infrastructure that is becoming critical in the wake of the ‘big data’ deluge.


The rental market In 2007 when Jason Stowe, CEO at Cycle Computing, first began to promote the notion of running technical computing on externalised virtualised hardware, he was met with a great deal of hesitation and scepticism. Tere are many reasons for this, which will be discussed later in this article, but one of the foremost reactions he encountered was resistance to the idea of running a workload on hundreds of cores. One of the company’s first customers had a workload that took six weeks to run internally. Cycle Computing ran it on 400 cores in a matter of 18 hours – once the client had been convinced that the use of the then enormous number of cores would not come with an equally large price tag. In actuality the computation cost for that particular workload was in the region of $400. Stowe also cites a more recent example,


where the company did a 10,600 server- instance run for a Top10 pharmaceutical company that involved 341,000 hours of computations over an 11-hour period. Market intelligence firm IDC estimated that to deploy a system capable of handling that workload would have cost $44 million in infrastructure and required 14,400 sq. ſt. of data centre space. Instead, Cycle Computing’s soſtware for creating HPC clusters in the cloud, CycleCloud Soſtware, was able to orchestrate it from Amazon’s AWS cloud in about two hours, and run the computation for an additional nine hours – all at a total cost of $4,372. Tis is the most attractive aspect of cloud computing; the ability to scale up and down infrastructure as required.


Times have changed in the past six years and Stowe believes we will see an increase in the number of companies that adopt a strategy whereby they no longer refresh their internal infrastructure, but instead will let it depreciate and start using cloud almost entirely. ‘Te pay-per-use cost option of cloud, coupled with the ability to change hardware dynamically to meet workloads, means that an increasing number of organisations are only deploying in the cloud or on internal/cloud hybrids. Of course, there are customers who continue to focus on the deployment of internal-only resources, but this mindset is quickly changing,’ he commented. Coraid’s Dave Kress agrees that server


virtualisation has disrupted our view of computing by allowing infrastructure to


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36