The Four Hundred

At the iSeries and AS/400 Connection Conference held last week in Naples, Florida, I sat down with Dr. Frank Soltis, IBM's iSeries chief scientist and the acknowledged "father of the AS/400." The visions he puts forth in the following interview are not necessarily planned items for IBM, but are predictions shaped by more than 30 years in the industry. No one receives universal or even strong agreement on forecasts, but few people possess the first-hand experiences of Soltis.

Computing is in its infancy--maybe adolescence is a better term--but it has not fully matured yet. When computing matures, I view it as becoming a utility. Some of the things we are doing at IBM and with the iSeries are setting that as a potential future.

Let me give an example of what I mean as a utility. Today, while I'm staying at this hotel, in order for me to connect with the information I need, I actually dial back to Rochester, Minnesota. I am talking to an iSeries sitting back in Rochester. It's the server that is supplying me with my gateway to the world. It all has to go back through Rochester.

If you think about that, it's ridiculous. Why can't I be using a server that's in the hotel where I'm staying? I'm not using my telephone network that is back in Rochester. I'm not using the power company in Rochester.

Why can't the hotel provide local service to me? Why can't the server in the hotel know that I'm here? That technology is there today in my cell phone. It should be there in my computer.

There has to be a worldwide backbone of servers. Those servers will all have to be in touch with one another. There will have to be some commonality, so that the server in this hotel will recognize various personal devices that will be used to communicate with the world.

When I move around to other geographic locations, and have access to servers as I travel, computing becomes a utility at that point.

I look at things like grid computing as a stepping stone to that type of environment.

Today we are talking about grid computing as tying multiple servers together and executing particular problems from a scientific standpoint. Let's look at it from a business standpoint. Grid computing can be used within a company's intranet. We can access some facilities or resources that are available on a server. In the future, I may want to use not only the server that the application is designed to run on but also resources from other servers. That's a direction to go when computers are capable of talking to one another.

The area of capacity on demand, at this moment, is a one-way street. And it's limited to a processor. As a customer of an iSeries you can turn on extra processors. Eventually it will be extended to memory, I/O, and other things.

In the future, you will not only be able to turn on the extra processors, you will be able to turn it off and give it back. As a business, you can recognize a heavy load--say, at the end of the month--that requires extra processors for three days. Not only will users be able to acquire and give back that processing power, but there will also be a way for billing for that service.

Another current example of where this is headed is that you can get WebSphere limited to a certain number of processors. In the past, when you wanted WebSphere to run on a box with eight processors, you paid eight times the license rate of a single processor.

With an iSeries, you can determine that WebSphere will only run on two processors, and therefore only pay for it to run on two processors. The mechanisms are built into the operating system to determine the number of processors being used with WebSphere.

It's like having accounting software in the operating system. To me, this is a stepping stone to the concept of utility. If, for instance, I am going to be using some aspect of the server in this hotel, the amount I use can be identified and I can be billed for that.

That's part of the reason why I am excited about being able to run different operating systems' applications on one server. As it is today, I can't assume that wherever I go the operating system will be the same as what I have. The convergence of operating systems will allow there to be servers that can run multiple environments.

This will require unprecedented cooperation among long-time competitors in our industry. Can you imagine that cooperation taking place?

You can say this is a grandiose scheme, with many problems getting all these companies to work together. But if you look at what's happening in the industry, a lot of this is occurring.

Economy of scale is driving it, to some degree. Look at the hardware business. It's very difficult to be in that business unless you have very, very large volume. As there becomes more and more hardware consolidation, potentially IBM could be building processors for many of its competitors.

Here's another hypothetical situation. Say that Sun Microsystems were to port Solaris to the Power platform and IBM agreed to support it, then you could not only run Solaris in a partition but also buy hardware from Sun and run an IBM operating system. I don't see that as a strange thing happening over the next five or 10 years.

As we start converging things, this concept of having a worldwide grid of computers becomes easier--they don't have to all come from one vendor. I'm a firm believer that if you want to have convergence, you have to start at the hardware level.

Hardware is the base, and then you need common operating system code to get the applications right. You have standards, and you start them from the bottom up. And that is the way I see the world going.

Some companies are going to say, "I'll just use this network." It's an ASP model. They will outsource everything. Other companies will prefer to keep their own servers.

One of the options for the company with its own server is to have that server tied into the grid, with its own unique part of the network. And, by the way, I believe it will be secure. Those companies can keep proprietary business information in certain partitions on the server and actually rent their excess capacity to the global network.

What can you expect in terms of these large companies working with one another?

Today we are a lot closer to that than we've ever been. First of all, look at the Linux phenomenon. I have no idea what Linux will be in terms of success five years from now, but I see tremendous potential. Linux is a cooperative venture involving many companies.

The business with the hardware is another. In the past, everyone just built their own. They could make it as incompatible and proprietary as they wanted. In today's world, the cost of that approach is tremendously expensive. The economics are forcing cooperation.

Is there a need to run Solaris on iSeries? Probably not in a business case. But the point is, there is more commonality than before. The industry has emerged to where much more is shared.

Suppose there is a worldwide network of servers, and wherever we travel we deal with local servers. Well, that local server had better be operating and had better be backed up. There has to be continuous availability. We will be dependent on it.

We also have to be able to guarantee security--company data and private data. We are not as secure today as we sometimes believe we are. The whole topic of security and privacy is an area that we haven't stepped up to as much as we need to. However, there are huge security efforts going on with various companies and government organizations, which will help drive progress in the areas of security and privacy.

A lot of the enablers are in place today. The best example is using the intranet--one server here and another across town. If I am running an application on one server and I need extra power across town, I will be able to tap into it. What is needed is high-speed communication to do this on a timely basis.

Some of the infrastructures are in place. In countries where they haven't had a wired infrastructure, they are skipping a generation and going directly to some of the high bandwidth kinds of things. The technologies are there, but they will be expensive to roll out.

Grid computing will first take shape within an organization. Eventually, as some of the protection and security features are added, it will open things to the outside of the organization.

Some pieces, like partitioning, are being used today. The capability of self-managing partitions. The interfaces to the partition structure, so that business partners can come in and do some automated management of those partitions, exist now.

There are a number of IBM business partners that are putting together packages that allow automated management of partitions. The next step is to manage from across separate systems, and the capability to expand or start a new partition on another system across town or wherever.

Within two years I bet there will be iSeries companies that are doing this type of thing. It's similar to the backup and recovery and switching disks that is being done today. I will be able to take part of an application running on one site and move it to another system with similar mechanisms. The automatic management of partitions will allow use of partial or separate applications.

It's sometimes frustrating, but the adoption of new technologies is slow. It's like when someone says, "Can you show me the iSeries customers that are doing business-critical applications on Linux?" It's a short list, but every month (or whatever time period you want to look at) it gets added to. Over the next couple of years it will start to show up.

Editor's Note: For more insights and opinions from Dr. Frank Soltis, see the article "Frank Soltis Talks About Shared Technology."