W3CHINA.ORG讨论区--Hey Leading Microsoft into Grid, HPC

http://www.gridtoday.com/grid/682714.html

Hey Leading Microsoft into Grid, HPC

By Derrick Harris, Editor

In this interview with Tony Hey, Microsoft's vice president of technical computing, he discusses a wide variety of topics, including: GridWorld Japan; the GGF/EGA merger; Web services reconciliation; and why he left his position heading the United Kingdom's e-Science Program to join the software giant.

---

GRIDtoday: You gave a keynote at GGF17 in Tokyo. What did you discuss in that speech?

TONY HEY: I discussed the need for the Grid community to use the opportunity of the merger of GGF with EGA to define an initial set of Grid standards the whole IT industry is happy to back. The new organization should have both an industry-dominated part -- which is tasked with producing relevant standards and tooling -- plus a research-dominated part that can propose and experiment with new specifications that may offer improvements in Grid functionality.

Gt: GGF17 was part of GridWorld Japan and, of course, we've already had one GridWorld in the United States with another scheduled for September. What do these events, especially with their frequency and global nature, say about the state of the Grid industry?

HEY: I think the GridWorld events demonstrate there is considerable interest and demand for Grid solutions from the commercial sector. However, as I said above, I think there is a need for GGF/EGA to deliver tangible results on a short time-scale that will give both Grid middleware developers and would-be Grid deployers confidence that their investments in Grids will be protected.

Gt: Microsoft has been cited on occasion as not being too involved with the Grid community. How accurate an assessment is that, and does your keynote, as well as the presentation by Todd Needham of Microsoft Research, signal a change on this front?

HEY: Todd Needham and Microsoft Research have been long time sponsors of GGF. It's true Microsoft has traditionally been focused on the business and consumer markets rather than on technical computing. With my appointment, nearly a year ago now, Microsoft is signaling its intention to engage more closely with the scientific and engineering communities. We will also soon have a Windows HPC Cluster product on the market and we wish to work with the HPC and Grid communities to better understand their requirements. Marvin Theimer of Microsoft's HPC Group is now actively engaged with the OGSA HPC Profile work. Grids, in the sense of utilizing "forests of clusters," are a natural evolution for Microsoft's HPC activity.

Gt: Speaking of the GGF, what do you think of the proposed merger between the GGF and the EGA? What kind of effects will the Grid community see from this partnership, and what kind of a role do you foresee Microsoft having in this combined organization?

HEY: As I mentioned before, I think the merger of GGF and EGA presents a key opportunity for the Grid community. There is now the possibility for the IT industry to unite on some simple standards for Web service grids that will benefit everyone. There is also the opportunity for the Grid research community to explore how best to extend these initial standards to provide more useful functionality. I think, perhaps, that in the past GGF has been too ambitious in its standardization activities. The new organization needs to clearly distinguish between building grids using tried-and-trusted Web services and experimenting with as-yet-untried, new Web services specifications. I hope Microsoft can work constructively both with other IT companies and with the research community to grow confidence in the new organization.

Gt: In more recent news, something for which you've been pushing is on the verge of happening with the reconciliation of the WSDM and WS-Management families of Web services specifications. What will this mean for Grid computing and its related technologies? What about the GGF with OGSA?

HEY: I am very pleased with the proposed roadmap for reconciliation in the system management space. I think this is good for the whole IT industry and for the Web services approach to service-oriented distributed computing. This reflects broad agreement on several basic specifications (primarily WS-Eventing, WS-Transfer and WS-Enumeration) and a roadmap to develop additional functionality above them. However, it will take some time for the Web services community to agree and to implement the roadmap. I therefore think it unwise for the Grid community to base any standards on the parts of the roadmap that have yet to be completed. I therefore very much support the approach of defining profiles of a small number of simple but extensible OGSA services that do not depend on future system management Web Services.

Gt: What do you think led to this decision, and why do you think it took as long as it did to be made?

HEY: There are many possible reasons, but one should not forget that the systems management space embraces a very wide range of "systems" -- right from the data center down to the silicon chip. So getting agreement in this space is a difficult thing to achieve and we should all be pleased at these signs of progress.

Gt: Now that you've been with Microsoft for a while (about a year), what do you see shaping in terms of the company's Grid strategy? Will we see Microsoft become a major player in Grid computing, and HPC in general, in the years to come?

HEY: We are taking our first step into HPC with our Microsoft Windows Compute Cluster Server (CCS) this year. In the next year or so, we will be enlarging our vision to include clusters of clusters -- or computing grids. At the same time, Jim Gray from Microsoft Research has been working closely with the astronomy community as they gain experience with their "Virtual Observatory" project. With their SkyServer and SkyQuery implementations, the community has a working example of a data grid built on basic Web services. So, we already have experience in some parts of the Grid space. In the future, I hope Microsoft will be seen as a major contributor to both Grid computing and HPC.

Gt: How important is it for ISVs and vendors to think about Windows when creating Grid applications and solutions, especially considering its prominence in enterprises?

HEY: I think all enterprises are concerned about TCO and I think they -- and the vendors and ISVs -- will indeed think very seriously about a Windows solution for their Grid and HPC requirements. For example, our HPC Windows CCS software is designed to easily take advantage of Microsoft's Active Directory identity management services that are commonly deployed in the enterprise environment.

Gt: Finally, I'm wondering if you could discuss what led to your decision to join Microsoft after spending years spearheading the UK e-Science Program. Was it just a natural move?

HEY: Strangely enough, I think it was a natural move. In both my previous roles as an academic research computer scientist in parallel computing and as leading a U.K. government initiative in e-Science, I have greatly enjoyed working with industry to solve real technical problems of relevance to them. It is now very exciting being on the other side of the fence working both with industry and with the academic community. I also think this is a good time to join Microsoft now that it is entering the technical computing arena with its first HPC product. With our commitment to interoperability and with standards such as Web services and OpenXML, I think there is the opportunity to demonstrate how Microsoft software and tools can co-exist and add value in a heterogeneous research environment that will typically include software from many vendors, open source software and Linux platforms. And there is also a connection back to my research into parallel computing. In the multi-core future outlined by Intel, clock speeds will not increase and we will have to exploit parallelism on the chip to gain significant performance improvements. This is a challenge for the whole IT industry, as well as for scientists and engineers.

Gt: Finally, how do you think your time spent with the e-science community plays into your ability to achieve your goals in your current position? Do any of the innovative projects the e-Science Program undertook mirror any of Microsoft's plans?

HEY: My time with the e-Science community gave me a unique insight into the next generation of scientific problems faced by the different communities. Almost every field -- from particle physics and astronomy to environmental science and bioinformatics -- will be faced with a veritable deluge of data from new high resolution surveys, from high throughput devices and from satellites and sensor networks. Rather than wasting the enthusiasm and talents of science graduate students by assigning them the task of building systems capable of handing, analyzing and mining literally petabytes of data, scientists should look to computer scientists and the IT companies to raise the level of abstraction and to provide them with the components of a reliable and functional cyberinfrastructure. And, yes, some of the UK e-Science projects were genuinely pioneering, and the lessons from these projects, and from other similarly ambitious projects in the United States and elsewhere, will certainly inform Microsoft's plans for technical computing.


	W 3 C h i n a ( since 2003 ) 旗下站点苏ICP备05006046号《全国人大常委会关于维护互联网安全的决定》《计算机信息网络国际联网安全保护管理办法》	46.875ms