Grid Computing: What I like about GOOP
GOOP is the acronym that Dan has used to talk about Grid Object Oriented Programming which is what the Digipede SDK supports. Soon I’ll be going out and spreading the word about Digipede and to prepare for that I’ve been researching the history of grid computing and the Globus project in particular. Globus shares information between compute nodes via protocols. This works great for heterogeneous systems and for the problems Globus is trying to solve. Globus is trying to do for resource sharing what the Internet has done for communication. And here is the big problem. The Internet was the highway, but the World Wide Web was the vehicle. It is the Internet that is protocol based and the Web which is API based. The Internet took off only when it became accessible via the APIs. Digipede has provided an object-oriented API for grid computing.
Now back to GOOP. Think about the number of programmers trained in object-oriented programming verses the number of programmers trained in network protocol development. The fact that programmers can use a well documented and understood methodology will allow them to grid-enable their applications quickly. The Digipede Framework SDK’s API works the way that programmers already work. This is important. Your programmers will be able to grid-enable existing applications faster because GOOP follows the same methodology.
Yes, I am very partial to the Digipede solution and it’s not because it’s my job. I really think this is a powerful paradigm shift that can have a fast and immediate positive affect on a company’s bottom line. The fact that you don’t have to buy servers to accommodate peek times but can roll that work over to existing desktops is an incredible cost saver. (We call this scaling-out.) The fact that the API is simple and follows OOP methodology means that you don’t have to hire expensive consultants or totally rewrite your applications. The fact that once your applications have been grid-enabled, computationally intensive processing times are reduced from hours to minutes. (Can you say increased productivity?)
Yea...I love GOOP.
Labels: Concurrency
posted by Kim Greenlee @ 10/25/2005 03:45:00 PM
4 Comments:
I'd like to see an article the discusses Digipede's solutions in terms of Ian Foster's seminal paper "What is The Grid?", because to me, this just looks like single enterprise, single platform, distributed computing solution, whose marketing department is misusing the grid name.
The Digipede solution is not for The Grid. I don’t believe that The Grid is commercially viable right now. I don’t think it will be either until people become comfortable with the both the engineering and business paradigm shifts. Businesses are building distributed solutions internally and they are calling them...grids. Whether it’s really distributed computing or parallel computing or grid computing doesn’t matter because the companies are using the term grid. Communication is always a negotiation process especially between people and since there is not one central agreed upon definition of grid, companies (like Digipede) will use the words that prospects and customers are using.
Now let’s talk about the value of a homogenous grid solution like the one that Digipede is providing. First, the “not invented” here philosophy is expensive. It is expensive with respect to both wasted resources and lost opportunities. Few companies have the money to commit to a large project with unproven technology. The Digipede solution is a low-cost, easy to use, middle-tier system that provides many of the benefits and few of the risks. By focusing on the Windows operating system, the Digipede Network allows companies to use the idle Windows compute cycles where ever they are found in the organization. The other value of focusing on one operating system is that few business applications run on more than one system.
Digipede is not trying to create The Grid. What the company and its products are trying to do is to solve real world problems for real people. And to let them solve those problems quickly and at a manageable price.
I'd add one point to Kim's. Digipede resisted the moniker "grid" for the first two years of its existence for the very reason that Anonymous points out--it's a single platform solution (although it has the communications and security infrastructure to exist across organizations).
However, we had a difficult time in the market using the term "distributed computing." Inevitably, customers would listen to a description, and say "Oh, you mean grid computing."
Every grid conference I have ever attended has spent significant time trying to define "grid." Is a cluster of clusters a grid? Is it a grid if it's just one OS? Is it a grid if it's in one organization, even if there are 1,000 machines on it? And, to get to marketing words: is Oracle's database a "grid?"
I'm familiar with Dr. Foster's paper. He defines grid as "a hardware and software infrastructure that provides dependable, consistent, pervasive, and inexpensive access to high-end computational capabilities." I like his definition because it gets to the heart of what grid/distributed computing is all about: aggregating compute power in order to increase capabilities.
In the end, there are many flavors of distributed computing. Grid is a handy (and shorter) moniker than "distributed." And while grid purists may disagree with our use of the word, we'll just have to stick with it as long as it's the one people recognize and understand.
Kim/Dan, you've hit the nail on the head. Whatever it is that is being called grid computing today, is not the grid computing that scientists and businesses are theorizing for tomorrow.
Dan's comment about resisting the word "grid" is most insightful and refreshing to read from a businessperson. It's understandable that to sell, you have to communicate, and to communicate more you have to use shorthand.
I've also attended presentations where we've stroked beards and discussed what is, or is not a grid - followed by Oracle presenting the 10g product as the Grid, period. We laughed at them all the way to the bar, which they paid for all night.
And therein lies the rub. Foster's “What is The Grid” paper is great because it defines 'grid computing', but the term has grown beyond (or was never really constrained by) this strict moniker because it's just easy to grasp, so people understand it, and can justify investing in it.
(I'm reminded that Hoover has similarly usurped vacuum cleaner, in general use, even though there is a specific definition.)
So, thanks both for your comments, they help put the whole grid thing into a good commercial perspective.
Post a Comment
<< Home