Grid Computing Explained

An anonymous reader writes "What's different between Grid computing and P2P, CORBA, cluster computing, and DCE? This article provides a cursory analysis of the similarities and differences between Grid computing and such distributed computing systems as P2P, CORBA, cluster computing, and DCE."

Comments

CORBA?

[TheRealMindChild]

CORBA doesn't even FIT into this article, and it is painfully clear when you read it. If this wasn't a free object model, I would say the autor is stretching to sell something, but since that can't be the case, then what is it?

Re:CORBA?

[neglige]

In a way, it does fit somehow (or so the author reasons):

Of all distributed computing environments, CORBA probably shares more surface-level similarities with grid computing than the others. This is due to the strategic relationship between grid computing and Web services in the Open Grid Services Architecture.

Oddly enough, WSDL and SOAP are mentioned, but never really discussed. And the would be probably better suited than CORBA.

Re:CORBA?

[roboros]

Oddly enough, WSDL and SOAP are mentioned, but never really discussed. And the would be probably better suited than CORBA.

WSDL and SOAP are used in version 3 of the Globus Toolkit, which is a well-known open source grid toolkit. In that version, all grid services are Web Services.

Indeed

[JMZero]

The author seems to have intuited their similarity by looking over their respective buzzword lists. All-in-all, the article seemed only vaguely in touch with reality. And it was targetted much more at management than technical folk, the kind of people who would ask "Should I be pushing my guys to use P2P or CORBA for the new CRM system?"

I'm fairly sure I'm stupider now than when I started reading.

The problem with Grids

[duffbeer703]

The big problem that I see with "Grid Computing" is that 99% of articles about it point out that it is cool and leave it at that. A few articles will point out how GM does rendering for CAD/CAM stuff or how Folding@Home works.

My question is what business problems can be solved with grids? Most people do not work in scientific computing facilities and most engineering departments are overseas anyway.

Re:The problem with Grids

[dkf]

Grid computing is still very new, and there is a lot of disagreement over what it means. Is it any wonder that the IBM guy just glosses over all that stuff to make a story that is journalistically stronger (but ultimately an utter crock, as you spotted.)

The distinguishing factor about Grids is that they operate across organizational boundaries. There's no point in doing a single organization Grid. Beowulf clusters? You might have them on the grid as single nodes.

BTW, there is one very well known (in the US) Grid system and there is one production-ready Grid system, and they are not the same.

Re:The problem with Grids

[Uma Thurman]

Grid computing means several things:

-the supply of compute power will equal the demand
-shifts in supply happen automatically
-you pay for what you use

Don't think of cluster computing. That's the wrong idea. Don't think engineering. Those people need clusters, not grids.

Grids are for people who run businesses. Their demand changes all the time, and it's expensive to have to buy computers that can handle the peak load. For example, some types of retailers will do 80% of their business in December. Why should they hang onto the computers all year around?

This is how it will work: you are a business. You call up IBM and you ask for a single processor Celeron machine, with 1 gig of database attached to it. They provide it, maintain it, make sure it's running, and charge you X dollars per month for it.

Then, one day, you discover 100 million dollars in sales that you have to get out the door. You call up IBM and tell them that instead of a dinky Celeron, you need a 32-way machine with a terabyte of database storage. They set you up and you've got the capacity on-line in 15 minutes. A machine like that might run over a million bucks to buy new, but you can use the machine for a lot less than that.

You use that capacity to get the orders out the door, then you discover that your business is back down to the celeron. You call up IBM and have them change you back.

Without grid, you would have to spend a million bucks or more to have that 32 way machine sitting around ready for the two weeks when you needed it. With grid, you only pay for two weeks usage of a million dollar machine, which is a huge savings.

Re:The problem with Grids

[Chris_Jefferson]

While that sounds good in principle, the problem is that for most companies the limiting fact is that they have too much information to deal with rather than not enough processing power.

Grid requires that you are doing things that don't require too much data to be pushed around (at least not compared to the amount of work that has to be done on the data). For large databases and the like, the problem is just sorting through the data and sending it somewhere else won't help because the effort of sending it there is probably no more than the effort of just looking through it for what you wanted.

Re:The problem with Grids

[DeepRedux]

On answer to the question of "what are Grids for" is given by the paper The Anatomy of the Grid. I think the paper can be summarized by the following quote from it:

The real and specific problem that underlies the Grid concept is coordinated resource sharing and problem solving in dynamic, multi-institutional virtual organizations. The sharing that we are concerned with is not primarily file exchange but rather direct access to computers, software, data, and other resources, as is required by a range of collaborative problem-solving and resource brokering strategies emerging in industry, science, and engineering. This sharing is, necessarily, highly controlled, with resource providers and consumers defining clearly and carefully just what is shared, who is allowed to share, and the conditions under which sharing occurs. A set of individuals and/or institutions defined by such sharing rules form what we call a virtual organization (VO).

Re:The problem with Grids

[Uma Thurman]

It's not just the CPU that scales in a grid environment. Things like network connectivity, and database storage also scale. If you haven't looked at the cost of some storage solutions from EMC, then you might not know how expensive they are. Renting can be a way to reduce cost, since these grid solutions can host many many customers on a single piece of physical hardware, without their customers knowing that they don't have their own dedicated box.

Re:The problem with Grids

[Jim_Maryland]

We are currently looking at grid computing for running CPU intense algorithms for preprocessing data streams. We are also looking at using P2P technologies for handling data between the systems (although this creates issues in our case for data backups - at least with our current implimentation). In our case, JXTA seems to be the appropriate route, but we are considering others as well.

Most people do not work in scientific computing facilities

True, most people are not involved in this sort of work. Over time, the developments in Grid computing may benefit other applications.

Re:The problem with Grids

[natmsincome.com]

Apart from raytracing the only other experiance I've had with grids is compiling using distcc. Apple has also brought out xcode which does the same thing. Basically think of anything that takes a long time and can be done in parallel. For compiling it make allot of sense. It means that instead of taking 1 hour and 1 minutes to compile it only takes 10 minutes. Add up how many times you compile the product and it adds up to alot of time that can be spent testing or debuging instead of waiting for the program to compile.

As for not many problems that can be solved using grids well how many things do you know of in this day and age that take longer than an hour to complete or require more than 3 computer(web server, web service, database) to run? If you don't know many/any they you won't really need grid computing either. It's not for you just like most people don't need a real time OS either but if you do then you don't have much choice.

Re:OT: Is there a Java-based P2P system?

[pcraven]

Well, the article was about grids, not p2p systems.

Here's one toolkit for creating grid programs with java.

Personally, I just don't see grid computing work where you ship your stuff out to 3rd party computers. There is the network latency, and the security aspects. But it might work for a company to maintain their own grid. That I could see. Maybe.