Breakfast with Roel

Computing Japan: Over the next decade, will Japan be a leader or a follower in the world networking market?

Roel Pieper: That depends on what you mean. If you make the differentiation between being a producer or being a buyer, you'll get a different answer. I think Japan as a technology producer is going to be a follower, because it has not invested that muc h in being a provider of networking products. In terms of being a user, though, Japan might go further and be quicker than what most people expect. We must assume that within a few years, most Japanese companies will have adequate networking solutions. It will be a surprise for a lot of people to see how quickly expectations change.

CJ: But isn't the Japanese networking user market currently lagging behind that in other countries, and isn't it predominantly a lower-end market right now?

Pieper: I think the market is lagging with respect to networking in general -- that's a true statement. But as companies net more and more comfortable in networking their computers, I think we'll find there's not a lot of difference between Japan, the US, and Europe. Same needs, same features, same functions. I would say that it is just a time issue, not a functionality issue. ATM can live in the local area network, and it can live in the wide area network.

CJ: Frame relay seems to be just taking off in Japan, but ATILI (asynchronous transfer mode) seems poised to quickly come onboard. Do you see Japan as leapfrogging over frame relay and going directly to ATM?

Pieper: I think we have to categorize a bit, decide in which kind of buckets we would need to put these technologies. The bucket that you put frame relay and ISDN (integrated services digital network) in is typically wide-area technology.

ATM is interesting because it can "cross the buckets." ATM can live in the local area network, and it can live in the wide area network. That is, of course, its promise: that one could start to look at a network fabric that is more uniform, from a desk top computer all the way out to a telephone computer to New York or London a fairly well-defined (what we call "flat") network.

CJ: So, do you think that frame relay and ISDN will be able to withstand the assault of ATM?

Pieper: In Japan, I think ISDN looks much stronger than anything else at this moment. NTT has been heavily sponsoring that technology. I believe that frame relay and ISDN are going to be popular both here and in the US.

ISDN is going to be especially interesting. We've got some new ISDN products for remote offices that would mean you could have a regional office -- insurance companies, banks -- and that office could run a local Ethernet local area network (LAN), and b asically use ISDN for a dial-up approach to the main office as the wide-area link. That looks like a very cost-effective technological solution.

CJ: Why has the ISDN solution been so slow to take off?

Pieper: The difficulty with ISDN is that it is very difficult to set up. The carriers who would provide that service need to do a lot of work to get such a communication system set up. Typically, however, their people are not very well trained, which c reates a big risk that the setup is not done right, that mistakes will be made.

The costs of setting up these links for the telephone companies is also likely to be much higher than for another capability. ISDN is at a disadvantage because it requires too much "manual intervention" for system definitions, session definitions. Not when you're using it, but when you're installing it. ISDN has a very maintenance-intensive installation definition process.

I think that both ISDN and frame relay will be at a disadvantage, over a five year period, against ATM -- both from a functionality point of view and a cost point of view. With any technology, though, there will never be one; there will always be many. You just need to know the priorities for the business view that you have.

We are demonstrating a system at Interop in Tokyo that actually has the choice of changing the wide area protocol. The user can say, "I'd like to use ISDN," or "I'd like to use frame relay," or "I'd like to use ATM." It's a modular wide-area network (W AN) switch. The product is called LightStream, and it's intriguing because it is not only modular at the wide area, it is modular at the local area. It sits at the boundary of the LAN and at the boundary of the WAN. A user can make choices on the LAN betw een Ethernet, token ring, and FDDI, and choices on the WAN between frame relay, ISDN, and ATM. It's a very flexible system.

CJ: So you think that, over the long term, ATILI holds promise as a viable corporate networking solution?

Pieper: If you look at the local area network world -- and this is not specific to Japan -- you're going to see a lot of Ethernet and switched Ethernet. The cost points, the price points, the capabilities of those technologies are good, stable, and ade quate for 90% or more of the applications today.

For special applications you need a higher speed, either a desktop connection or backbone connection. Today, with ATM, you would probably go to a CDDI- or FDDI-type connection, but the cost points and the complexity points of those solutions are fair ly high. So, FDDI-XT is being attacked by switched Ethernet as a backbone solution, and it is being componented with backbone router solutions that create a shortcut. Without going through a funnel, they basically create direct links to the client world a nd the server world that you're trying to connect.

CJ: How about routing as a long-term corporate solution?

Pieper: The problem, as a company, is that if you're using routing as a kind of real wide-area connection or backbone connection model, you'll eventually wind up with so many routers that they become totally unmanageable. Some companies that have gone to routing are basically stopping, and wondering what they should do now, because the management and maintenance of such an environment is very, very complex.

Specific, dedicated router links are feasible for companies with two or three subsidiaries or remote offices, but if you start thinking about companies trying to communicate with a hundred subsidiaries and remote offices, it soon becomes a real mess. T hat's why I think ATM is going to be so interesting. ATM provides a switched architecture by which you don't have to predefine who you're going to call, and how. The system will try to figure it out by putting a logical layer over the physical connection fabric; it will find a path that makes sense. That's the fundamental advantage of ATM in the wide area. And in the local area, the consideration is going to be price. The bandwidth will be so overwhelming that the decision will be based on how low you can make the price point for ATM.

CJ: Isn't ATM primarily a high functionality, high-speed protocol?

Pieper: ATM is a protocol that can run at different speeds. It is a mistake to think that ATM has to be high speed, You could do wireless ATM. ATM is only a logical protocol. Obviously, the lower the speed, the less the advantage. It's like RISC comput er architectures; the advantage of the RISC instruction set is that at very high CPU speed, you have almost linear behavior for instruction execution. At very low speed, the advantage of the RISC instruction set becomes zero, or negative.

In networking, it's a similar situation. The advantage of ATM is that it can be processed at incredibly high switched speeds, because of its fixed cell architecture; so its protocol management is fairly simple. If you move the speed down, other proto cols start to look more interesting and cost effective. You could do ATM protocols for cellular phones; it's technically very possible, but it probably would be much more expensive than the current system. You can do wireless LANs with ATM; the question a gain is functionality, speed, cost....

CJ: In terms of speed, how low can ATM go?

Pieper: The lower end for ATM is close to 10 Mbits, but I believe the more likely speeds are 155 or over. Some companies have announced 25 and 52, but T think that is a mistake, because the cost-points of those solutions are going to be fairly close to the 155. I believe 155 will be a major backbone and a good desktop speed. Still, the cost of ATM to the desktop for large networks is going to be prohibitive for a while, and many people will go with switched Ethernet.

CJ: If you were setting up a network for a medium-sized company, what would be your preferred solution?

Pieper: If I were doing a network today, as a user, I most likely would use ATM backbone 155, and switched Ethernet and Ethernet for all desktop connections. I feel that 155 makes a lot of sense because most wide-area speeds will be 155. So you would h ave the same blocking factor -- or nonblocking factor -- from the desktop all the way through the network.

If you have 155 in the wide area and much slower speed in the local area, you need a computer that can block/deblock, multiplex/demultiplex these communication systems, to different speeds. The advantage of having all the same speed is that it becomes even simpler to manage -- less congestion management, less processing in the network. For now, I think that 155 for the user groups that need it will be an interesting solution, probably the simplest one that works. If you need to block and deblock 25 Mbi t to 155, for a lot of users, it's going to be another technology challenge.

We need to get to dumb networks, where very little management takes place in the network; most of it takes place on the outside. Cost-effective, dumb networks is the model for the future.

CJ: Do you have any advice for companies considering network design and implementation?

Pieper: One thing I think your readers should give some thought to: One of the hurdles with regard to networking is that, to really take advantage of networking PCs and networking people, the company's organizational discipline and the organizational m odel is crucial. If you don't, as a company, realize the need to change the way you make decisions and how your business processes work -- if you have a very hierarchical organizational and management structure -- then networking your company might not be very useful. All the efficiency advantages that you could achieve are not going to happen if you're still pushing the paper up and down the ranks. In that case, networking is going to be a useless investment.

Another very critical point is that a lot of these people are simply connecting computers together, but they don't really understand what communication data model must be created so that the data provided to those environments is correct and current. The work necessary to ensure the proper information flow to these networks, and the need for the files to be current and accurate, is a big problem. A lot of people make copies of files and never ask: "Is the data still valid?" "Is it still up-to-date?" T hey're doing all this stuff from totally obsolete data files. The need for the information model to be in line with what you're trying to achieve, and to ensure that the data is current, is a big problem. But a lot of companies don't have that realization yet.

Many companies will find themselves having networked and then suddenly face another hurdle. If you have a thousand employees, you might suddenly find you have 800 copies of the same file -- almost the same, but all different and all out of date. The wh ole information design process for your data, and how the data is connected in being current to your business process, your ordering, your manufacturing, your distribution -- that's quite a challenge.

CJ: Those are two very big hurdles you don't hear people talking about very much.

Pieper: We've gone through a very big reorganization ourselves, and I actually forced those problems on our company. We restructured, took out all middle management; we went to full implementation of a Lotus Notes network; we created business applicati ons for customer support, customer service, order entry, manufacturing, distribution....

And guess what -- the first thing that happened was the data got out of date, out of synch. The customer data wasn't right, the financial data wasn't right. So we've lived through these problems. We restructured: we went from 1700 people to 1100 people . We created a very flat organization, we created very focused development teams, business teams, marketing teams, manufacturing teams -- all connected on the same network with their own value-added applications and what we call our value-added databases behind it.

We are just starting to get out of the real problem phases that were caused by not having the proper information model, not educating the people.... So, we are an interesting role model ourselves of what this is all about. I spent a lot of energy mysel f on making sure that I can represent what we've done as, "this is one of the ways it could work."

CJ: It must be helpful to tell customers that you've done it yourself, so you know the problems that they will face

Pieper: Yes, and not just talk about it from a technology/cabling perspective, but rather from what application stuff is involved, what data is involved, all the education processes involved -- it's very, very underestimated. It's totally underestimate d here in Japan.

CJ: What is the UB Networks strategy for succeeding in the Japanese market?

Pieper: Many companies here, like Fujitsu and NEC, operate with both a product strategy and a systems integration strategy. And that is actually our worldwide business model. In a customer or competitive situation, we don't lead with product issues; we lead with the ability to put the network together. Technology is important, but for us it comes second.

Networking is quite a new segment in Japan, but having been here for 12 years, our experience base is quite unique. You can buy a product, but you cannot buy knowledge, you cannot buy experience. Our experience is an asset we hope to leverage, both wor king directly with customers or indirectly through distribution partners or systems integration partners.