Tampere

Friday morning, I headed north to Tampere in Finland. With three hours to kill in the Helsinki airport before my plane left, I idled away the time trying to decide between the sauteed reindeer with lingonberries and the elk pate with rowanberry gelatin. It was 4 P.M. and the sun had just set.

I was met at the Tampere airport by Vesa Parkkari, a man I often refer to as the Finnish Bill Gates. Of course, Vesa has a much more effusive personality than Bill Gates, although I confess my only contact with Mr. Bill has been through interviews in the trade press, which could make anybody look boring.

The next morning, I got up early enough to watch the sun rise at 7:30, then had a leisurely breakfast of pickled herring and brown bread. It was Saturday morning and evidently some sort of holiday in Finland. Of course, nobody had any idea what the holiday was for, but it was enough to close most of the shops in town.

It wasn't enough, though, to slow down Vesa. At 10 A.M., he brought me down to the company he founded, Mikrokonsultit Oy. The name is a bit of a misnomer, as they didn't have anything to do with microcomputers and didn't do any consulting. (Since my visit, they changed the name of the company to Relevantum Oy.)

In the late 1970s, however, Vesa was heavily involved with both microcomputers and consulting. He took early Motorola chips, made them into microcomputers, and then used the systems to design custom applications. One system was developed by Vesa for NESTE, the state monopoly oil company. Involving 2 square meters of electronics (and lots and lots of assembly language programs), the system coordinated the weighing of all lorries coming into the refinery to load oil, printing all the necessary lists and other documents.

Another, system, developed by Raimo Harju, his copartner, was still being used. The system takes raw timber and figures out the most efficient way to saw up the log to produce the most lumber. In an economy where 30 to 40 percent of GNP is based on forestry, this system proved quite popular.

Vesa had the same problem with consulting that others all over the world face. He did the programming, and thus bore the risk when a job took an order of magnitude more time to complete than planned. A fact of consulting, however, is that you often end up several subcontracts deep in order to get a job. Vesa bore the risks, but others higher in the food chain were making all the money.

Vesa decided instead to give seminars about putting microcomputers together. He gave a few at the Tampere Institute of Technology, and the demand quickly overwhelmed the university facilities. This was obviously a worthwhile business to be in.

Vesa and his colleague Raimo Harju went into business. At first there was no company, just a couple of guys in a hotel conference room. The hotel couldn't handle the idea of a meeting without a company name, so they made one up: Mikroboys.

Mikroboys became the more dignified Mikrokonsultit and business flourished. In the early 80s, however, the prepackaged PC came into being and Vesa saw that there would be a limited future for very low-level courses like his. In 1982, they switched the company to an exclusive focus on data communications seminars.

Seminars are a funny business. Almost everything is a fixed cost. You spend a month or two (or more) to develop the seminar. Then you print and mail brochures. Then you wait.

If enough people sign up, usually something like 15 to 20, you break even. Everything after that is gravy. Well, in 1983, the gravy began for Vesa. Finland had just started up two different public X.25 networks and Mikrokonsultit had a hands-on practical X.25 course. They began teaching seminars to 70 people or more on a regular basis.

Based on the early success, the company expanded and now has 13 different seminars. Finland, with a population of 5 million people, is a small country. Yet it is big enough to support Mikrokonsultit.

There are two keys to the early X.25 seminar success, and both factors hold for the later courses. First, courses are based on technology that is available and cost effective. In the X.25 realm, for example, Finland has two public data networks in heavy competition with each other supporting over 3,000 hosts. Leased lines are cheap enough in Finland that it is estimated that there are another 3,000 hosts on private X.25 networks.

The second key to Mikrokonsultit's success is that the seminars are hands-on. In a 5-day X.25 course, for example, users not only learn the theory of X.25 but how to configure Packet Assembler/Disassemblers (PADs) and hosts and how to run applications.

Vesa took me through some of the exercises that the instructor does in the X.25 class. We took a PC X.25 card, installed it, configured it, then loaded up the X.25 and PAD software. We then set up links to Russia, China, the U.S., and the United Arab Emirates. We looked at a network analyzer to analyze PAD and X.25 parameters. We looked at tariff tables and calculated breakeven points for leased lines versus packet networks. We ran SNA 3270 and X.400 over X.25 to real hosts.

Other courses on subjects such as X.400, TCP/IP, internetworks and routers, and SNA all take the same hands-on approach. Dragging Cisco routers, PCs, analyzers, and other equipment along means that you can't just slap an instructor on a plane and have him read the words on slides to the class. You need people who can both teach and use the technology and, in the case of Mikrokonsultit, are fluent in Finnish.

All day Vesa led me from one room to another and from one topic to another.

"Now we will do X.25 demonstration."

"Now we will discuss DECnet Phase V."

"Now you can use the Internet to read your mail."

At the end of the day, with my eyes glazing over, he had a new assignment.

"Now we will go to my castle and do Finnish sauna.

The training market had been kind to Vesa and he had recently bought a castle. Much as I wanted to see this castle, I was torn. I had visions of being thrown naked out of a scalding sauna into a moat filled with ice.

We arrived at the Parkkari castle, and Vesa showed me around from room to room. While his wife Pirjo Sistola and their three year old daughter took their sauna, we polished off a bottle of champagne. Finally, the fateful moment arrived.

"Now we will do Finnish sauna."

I tried to demur, but Vesa made it quite clear that failing to participate would be considered a grave form of national insult. After a shower, I entered a very, very hot room.

Vesa picked up a bucket and threw ladles of water onto the stones, producing blasts of steam. What I had thought of as very, very hot became an order of magnitude hotter. While Vesa kept throwing ladles of water onto the rocks, he explained the seven levels of the Finnish Sauna Reference Model.

I was experiencing level 1. Level 2 seemed to have something to do with more water on the coals. In level 3, you exit the sauna periodically and rub ice cubes over your body, then jump back in. Level 4 consists of whipping yourself with strips of birch bark to stimulate the skin. In level 5, you leave the sauna and run outside to the lake. Level 6 has you roll around in the snow. Level 7 was some form of Finnish national secret and, quite frankly, I didn't really feel I needed to know.


Sunday morning was spent discussing the Internet with Vesa's staff. We speculated on the motives of ANS and the fate of the NREN, after which I spent the afternoon taking a walk around the Tampere lake.

That evening, I was met by Juha Heinänen, one of the creators of the Finnish Internet and an important technical influence for the whole Nordic region. Juha took me to "Salud," a Spanish restaurant, where we feasted on a carpaccio of wild boar and an entree of alligator served with a coconut and papaya sauce.

Juha began setting up WCP links in Finland in 1982. At the time, Finland had a large X.25 network for connecting the University VAXen, and UUCP was used to transfer mail and news over phone lines within the computer science community.

In 1984, having finished his doctorate, he went to the University of Southwestern Louisiana for a postdoctoral fellowship. While there, he helped set up a TCP/IP network for Louisiana and, as he tells it, enjoyed many Cajun parties.

Coming back to Tampere, he spent four years as an Associate Professor at the Tampere University of Technology. At the University, he was asked to join the steering committee for FUNET, the Finnish University Network.

There was a problem, though. FUNET was basically DECnet running on X.25 and Juha had a Sun workstation. He persuaded FUNET that their role should be to connect LANs together, not VAXen.

At the time, there were no multiprotocol routers, so FUNET became a network based on wide-area bridges. A year later, however, Juha went to a USENIX conference and saw a couple of guys standing behind a small table.

This was the Cisco booth and the company ha just added DECnet support to their boxes, but couldn't find a distributor. He convinced a salesman at a local company to start carrying Cisco and two months later FUNET had become a multiprotocol network based on routers.

After his success with the university network, Juha became a consultant. His clients included FUNET, as well as Telecom Finland, the country's nominal PTT.

Finland, unlike most European countries, has considerable competition in telecommunications. At one point, in fact, there were over 700 telephone companies. In 1991, there were over 50, each serving a small area. In addition, there was Telecom Finland.

Telecom Finland was state owned when I visited, but was soon to be incorporated. It had two functions. It served as the local telephone company in areas where there were no others. It also functioned as the long distance telephone company. Even in the long distance market, it already had competition in data and would soon have so for voice traffic.

In all cases, Telecom Finland received no special treatment and competition was encouraged by the Ministry. Data networking was especially competitive. Finland had two commercial IP networks: Datanet, run by Telecom Finland, and Lanlink, run by a consortium of the smaller companies.

Both Datanet and Lanlink had connections to FUNET. FUNET, in turn, was part of NORDUnet, the regional academic network linking five Nordic countries. Datanet had another link into Infolan, an international IP service owned by a consortium of PTTs. Infolan, in turn, was connected to PSI's commercial network in the U.S.

All told, Finland had over 10,000 Internet hosts. This was 10 percent of all European hosts on the Internet, far in excess of Finland's proportion of the European population.

Both Datanet and Lanlink differed from their commercial American cousins in a very important respect. PSInet, CERFnet, and other U.S. networks were open networks. When you joined, you could talk to any other IP host. Of course, that host might not want to talk to you, but the backbone delivered packets up to the customer router or host.

In the Finnish commercial networks, access was explicitly controlled in the backbone. When a customer started up service they explicitly identified a set of hosts. Routers at the edge of the backbone enforced the policy with access control lists.

Datanet and Lanlink were thus more than just IP networks. They actually formed a sort of multiprotocol, LAN interconnection service. When I went to talk to Juha, Datanet was in transition. It was just beginning to provide the generic LAN interconnection service in a different way.

The old model consisted of a border router at the edge of the backbone, which in turn was connected to a router at the customer site. The backbone itself consisted of a large number of E1 lines running at 2 Mbps. As Juha explained it, "no need to scrimp on backbone capacity. After all, we are the telephone company."

The new model of Datanet was based on Frame Relay. The Cisco router at the customer site, equipped with a Frame Relay interface, established a permanent virtual circuit (PVC) to another router on the edge of the network. If a user wanted to send TCP/IP traffic onto the Internet, for example, a permanent virtual circuit would be set up to either the FUNET or Infolan router. Traffic from Novell, DECnet, or any other protocol could use the same PVC to send data.

Juha saw a number of advantages in moving the public network down from the network layer into a level 2 data-link service. First, bridges can be supported on a multi-user network. Bridges broad cast all packets to the other side. Since Datanet supports multiple private virtual networks, in the old model there was always the question of where to broadcast the packets. With permanent virtual circuits, the customer answers that question.

Subnetworks are equally difficult to support in a general-purpose Internet, like the old model. If a customer had a single class B address split over multiple sites, an IP-based internetwork would not be able to figure where to route the packets.

There is an even more difficult problem with a level 3 network, the question of policy routing. Let's say that there are two Datanet customers, one commercial and one a researcher, wishing to set up a connection to the same destination in the U.S. The researcher would want to use the FIJNET route and, in theory, the commercial user would need to go over the Infolan path.

When a packet from the first user comes in, a route to the destination would be cached in a router. Subsequent packets, regardless of the source address, would all go by the same route. Under the Frame Relay model, the user would set up a permanent virtual circuit to the edge of either FUNET or Infolan. Note that the policy routing dilemma may show up further down the road, but as far as the local (or national) data network is concerned, policy routing decisions have been eliminated.

Policy routing protocols are one of the most difficult issues in the Internet. Many are advocating the solution that Juha chose of simplifying the network layer topology in favor of a general, public data link service such as Frame Relay or SMDS. Others would argue that moving the public network down to the data link solves problems today, but is a solution that will not scale well as the scope of the network increases.

For Juha's employers, Frame Relay certainly solved several immediate needs. It avoided the policy routing dilemma for TCP/IP, it supported a variety of non-TCP protocols over the same infrastructure, and it did away with access control lists.

Monday morning, I stopped by Juha's laboratory at Telecom Finland to see the network and read my mail. Finland is so well connected that I got better response time reading my mail on a Telnet session back to Colorado than I got when I went in locally over a 9,600 bps dialup line. In fact, Vesa Parkkari had demonstrated FTAM file transfers on top of TP4 and the CLNS providing end-to-end performance from California to Tampere of 38,000 bytes per second.

I then spent the rest of Monday making my way from Tampere to Helsinki to London to Paris, with the obligatory many-hour layover at London Heathrow. It was a relief to arrive at my familiar hotel in Paris.

<<<>>>