The Union of HPC, the Cloud, and Performance: A Way Forward
The name of this publication, “HPCintheCloud” reminds us that the emphasis is not solely on applications within cloud computing environments. We must not forget about the relevance and focus on HPC in our discussions, as well as, the challenges and the ability of the cloud to meet those challenges. The advent of many- and multi-core technologies has dramatically enhanced our ability to achieve certain performance benchmarks given the challenges with heat, energy and power. Our obsession with hardware, however, has unfortunately run us into the brick wall reality of the limitations that we face if we try to go the road of hardware alone, and don’t include the role and benefit of software.
Hardware, as we have been recently forced to admit will only get us so far in our quest to make significant improvements in performance. A U.S. Council on Competitiveness report in 2005-2006 on HPC noted the peril and consequences that we should expect if software continued to be treated as an afterthought in our efforts to expand HPC usage in critical areas of technology need . Hence, the role of software and algorithms that enable users to take better advantage of the hardware advancements in general, and processor advancements, in particular can no longer be ignored .
The following questions (and others that we hope that you will soon share that you believe should be incorporated) will be used as an initial starting point in our discussions:
• Can we develop a “best practices” blueprint that can be applied across several classes of scientific applications?
• Can we expect a transition towards a more universal interface that is as commonplace and user-friendly as those being witnessed with the cellular phone industry?
• Can we achieve a secure solution that does not significantly minimize the flexibility that users need and desire at the expense of data protection?
• Can the discussion and relevance be extended to other “external” disciplines, such as in Fine Arts, Communications, and the Social Sciences? The advances in computing and information technology, along with our focus on data enables us to address a broad range of scientific applications that are not limited to the traditional national sciences communities of Science, Engineering and Mathematics.
• What’s Green Enough? In the battle to bring under control the heat, energy, and power dilemma on our resources and environments, a number of “Green Computing” efforts are being incorporated that could seriously affect the way HPC applications are engaged. Scientific applications, which are typically very data-intensive, as one would expect, are likely to put much more demand on the infrastructure and resources.
We look forward to our journey down a new road in pursuit of our familiar goal of moving a critical area of technology forward in a way that is not only mutually beneficial but carries the probable risk of being a lot of fun. Let’s roll up our sleeves and have at it. Cheers!
The invitation to be a contributing Editor to the new publication, HPCintheCloud, is quite an honor, for which the author is truly grateful to Tom Tabor at Tabor Communications.
 U.S. Council on Competitiveness Report, 2005-2006 on High Performance Computing.
 Dongarra, J. “”An Overview of HPC and Challenges for the Future”-HPC Asia 2009 Keynote Speech