Another Monday, another issue of GRIDtoday -- but this one is different. Why? Because it marks the inaugural article
from new contributing author Robert Cohen. And what an article it is.
Cohen takes a look at how Grid computing has made it mark in the automotive industry, especially as it related to the concept of "virtual cars," and coins yet another new term: collaboration grids. I'll let you find out the details yourself by reading the article, but this seems like the type of application about which the user community at large needs to hear. The concept of virtual products being designed and tinkered with by teams around the world via collaboration grids delivers on many of the selling points of Grid computing, and it seems logical to me that many other industries could benefit from this type of grid usage. Flexibility, global collaboration, time-to-market, ROI -- all the usual suspects are there. I'm glad Cohen was able to shed some light on this very interesting use of Grid, and I'm excited to see what topics he covers in the months to come.
In the same vein as Cohen's article, we also have a Q&A
(from this month's Globus Consortium Journal
) with Ravi Subramaniam of Intel. Although he doesn't go into great detail, Subramaniam gives a nice peek into how Intel uses its 60,000-plus node grid for the company's engineering design automation (EDA) work. Maybe it's just me, but I often times forget that many of the vendors (especially the hardware vendors) involved in the Grid space are themselves grid users. However, when I'm reminded (as I was when I first read this Q&A), it raises in me the same question as every other time I hear about a large corporation (be it Intel, AMD or eBay) using Grid technologies: What is everyone else waiting for? Intel and AMD, for example, are not like tobacco companies hawking cigarettes while their executives wouldn't dare smoke them - - these companies are actually using the technologies they are touting.
As for the rest of the issue, here are several other articles that might be of importance to you: Defense Agency Grants HP $440M Utility Computing Contract
; United Devices, Satyam Grid-Enable SAP Instances
; Harvard Adds Blue Gene/L System to Crimson Grid
; U of Tennessee Selects Ciena for Optical Grid Network Research
; and Sun Expands Virtualization Offerings
Finally, I'd like to point out that Oracle OpenWorld
is going on this week in San Francisco, and Grid-wise, one of the highlights should be a panel session including participants from AMD, HP, IBM, Oracle and Sun Microsystems. I got a chance to speak with Margaret Lewis, AMD's director of commercial solutions, about this panel, as well as about what's new on the processor front as it relates to Grid computing and virtualization. Lewis said the panel discussion came to be because AMD had hosted a similar one on a smaller scale, and Oracle asked them to do another at OpenWorld. The topics to be covered, she said, include: "definitions of Grid vs. virtualization," "where are users applying Grid," "how can technology handle business issues," "where is Grid moving," etc.
And could there be a better group of vendors to discuss such issues? Lewis had nothing but praise for the other vendors with whom she will share the stage, noting that IBM, Sun and HP all have their Grid roots in the HPC arena, and they're all taking this experience to apply Grid to business issues. As for the event host, Lewis had this to say: "Oracle has a unique perspective because Oracle considered Grid computing for business applications long before many other [companies] did, and started putting the whole concept of Grid computing out in terms of talking about it as a platform for delivering database and business applications."
As for the bottom line, and why the OpenWorld audience should attend this session, Lewis said it is a good opportunity to see how different vendors are approaching Grid computing in different ways, as well as where they're finding common ground. Interestingly, said Lewis, even though [the panel participants] come from different backgrounds, the evolution of Grid into the business realm has led to a lot of similarity among their various strategies.
On the broader subject of the relationship between processors and Grid/virtualized environments, Lewis gave me a nice background. Essentially, she said, "The processor provides the underlying functionality for a lot of what happens on the system." Today's Grid environments are increasingly CPU- and memory-intensive, she noted, and the latest generation of processors is definitely addressing these issues. With multi-core chips, embedded memory and hardware hooks for virtualization, processors are being designed to maximize Grid and virtualized environments.
Strict performance enhancements aside, Lewis said that new processors are also being designed to be more energy-efficient -- something that becomes very important in large Grid or cluster environments. This can be evidenced, she added, by looking at Google's current to build a solar-powered data center. Further, multi-core (which will be stepped up to quad-core in 2007) processors address other Grid issues, especially the ability to fit maximum CPU capabilities in a small space.
However, Lewis' most interesting point (from my perspective, at least) was her acknowledgement that while grids appear to be splitting into two distinct types -- computational and transactional -- AMD is seeing some commonality emerging between them. In terms of virtualization, she said, "A lot of customers have gotten into the mode with x86 servers where they do one application per server, and they solve their computational issues by just buying more servers. The result is they have a lot of underutilized servers." With the advent of virtualization, though, these customers are now trying to figure how to put multiple jobs on one server by consolidating virtualized servers together, where they can move VMs or jobs from one server to another.
To wrap up this point, Lewis noted: "On the transactional side, we see virtualization is being a step in for people to understand and start running grids effectively. On the computational side, people have been racking the computers up into clusters or grids and they're looking at how to manage them more effectively or schedule them more effectively."
And, as you might assume, Lewis believes the processor is very important as customers try to evolve and modernize their data centers -- perhaps more so than many people think. "It isn't like [the processor] is just a generic piece of silicon," she said. "For something as rigorous as Grid or virtualization, it becomes important."