云计算英文资料grid-article.pdfVIP

The Grid, Cloud Computing and our Manycore Future Dennis Gannon School of Inf ormatics Indiana University The Grid concept originated with a brilliant vision. Larry Smarr was the first to articulate the idea of computing as a utility that operated like the electrical power grid. You could “plug in” anywhere and information and computational services were instantly at your disposal. When we use the electrical power grid, we don ’t concern ourselves with which power generator provides us the electricity. We j ust use it. The computational and data Grid should operate the same way. For information services, the Web already operates this way. We don ’t care which of Google’s compute farms processed our queries or where the data I down-loaded was stored . But what about computational services? The Grid research community has been trying to build the standards to make Smarr’s vision a reality for nine years, but this year we have been confronted with some ideas that are altering the landscape. In fact, 2007 may be remembered as the year the future for Grids became very “cloudy”. The industry giants (Amazon, Google, IBM, Microsoft, Yahoo) have been building massive data centers at a rate that is completely outpacing the growth of the national supercomputing centers. These compute and data clouds are transforming our vision of what a Grid can be. Google led the way by showing how we can a use a large cluster of machines to do massively parallel data analysis tasks by uploading analysis codes into a ``MapReduce’’ software framework. Amazon demonstrated how a computing cloud could upload a virtual machine image to run arbitrary user applications as web services. Yahoo has made their computing cloud available to students to do further research on the MapReduce model. Microsoft has announced a billion dollar inve