Wolfgang Gentzsch says grid computing will come in three waves. The first, well under way, primarily involves the academic research community. The second, just beginning, brings in corporations as users. The third, still some years off, will add individual consumers to the grid. At that point, the Internet will be "the grid," says Gentzsch, managing director of grid computing and networking services at US research establishment MCNC. A preview of the grid can be glimpsed in the National LambdaRail testbed network, for which MCNC will serve as a technical support centre.
Gentzsch recently told Gary H. Anthes what it will take to ride the waves.
Q: What's moving us into the second, corporate, wave of grid computing?
A: The IT vendors have their grid story in place - IBM, Sun, Oracle and the others - for the next generation of products that they want to ship and make money with. But there's no money in research grids and consumer grids are far out. So the current interest is in the enterprise grid.
Q: But are users interested?
A: Grid technology can improve equipment utilisation from 30 per cent on average to something like 70 per cent to 80 per cent. So the interest of end-user companies has grown dramatically. Driven by the benefits - easy access, sharing of resources, easy fail-over and dramatically improving productivity - within the near future companies will reduce cost and complexity [with grids].
Our products and systems are getting more and more complex and the infrastructure is getting more complex. Scaling the infrastructure means scaling the people. But if you have a grid in place, you can handle a factor-of-10 increase in resources with the same people.
Q: How pervasive will grids be in corporations in, say, five years?
Companies in the life sciences and the financial sector are already advanced today in using local grid technology. In about five years, many of the Fortune 500 companies will be still growing globally, having subsidiaries all over the world, and within these subsidiaries they have IT equipment that may be empty because it's night. So now in a global enterprise grid, a company in Europe, for example, might want to send jobs to America, where people are still sleeping.
Q: Life sciences and financial companies tend to use grids for scientific applications and modelling. What about mainstream transaction-processing apps?
A: You can't really easily send independent transactions over the Internet. Sending a transaction to a remote server takes far too much overhead, so you send a whole application to a transaction server. You have several transaction systems distributed. It's a grid because the resources are distributed.
Q: Given all these benefits, why aren't we seeing much more rapid adoption of grids?
A: Let's take the hype out of it. Grids promise much more than they can deliver today because grid technology is not mature and complete. There are point products which solve a little bit but don't transform your whole environment into a grid. Grid standards are still missing.
And there are challenges beyond the technology. Server-hugging is still a key barrier to grids; people love their own little systems. Security is another issue. Also intellectual property - should I really run my IP-heavy stuff on another company's server, or send it over the Internet? There are legal issues, social issues, political issues.
Q: What's missing in standards?
A: The Global Grid Forum is working on these issues with something like 50 working and research groups in seven areas [such as applications, architecture, data, security and performance], and basically all the areas are immature.
For example, how to grid-enable software applications. Today you have to write interfaces manually. These interfaces need to be standardised so you can plug two software modules together seamlessly and at all levels - application, architecture, data, security and scheduling and resource management.
Q: You speak of the grid and the future Internet as if they were one thing.
A: The current Internet is just information. The grid adds things like collaboration, computing and other things that make it three-dimensional. It's doing real business and development and production on top of that. With [today's] Internet, you have information. But with the grid, you have access to the resources that use this information. Grids are getting more and more general-purpose. Eventually they will merge, and you have the next-generation Internet.
Q: Will it be based on the same standards as today?
A: Current protocols, like in SONET, Ethernet and TCP/IP, were built more than 20 years ago, and they never anticipated the amount of data which they have to deal with today ... for the "grand challenge" applications. That's why the National LambdaRail has been started, and its key focus is to overcome the limitations of TCP/IP and other protocols.
Q: What about the third grid wave, the one for consumers?
A: We are talking about gaming grids, where hundreds of gamers come together and use the grid for really heavy interactive and compute-intensive stuff.
Also health care. If you have a heart attack or stroke and you are within 15 minutes of a hospital, you get easy help. But in the countryside, the percentage of people dying from heart attacks is at least 50 per cent higher than in the cities. Now, a grid reduces distances to zero, so the country doctor has immediate access to all these expensive machines, which have digital heartbeats, in the hospital. If that hospital is too busy, the health care grid broker selects another resource that is least loaded.
Q: How far off are these third-wave capabilities?
A: Five to 10 years. Five is aggressive; 10 is definitely yes.