'X' marks the spot at MIT: Why IBM and DEC have contributed millions to shape Athena

Paul R. Strauss
Data Communications

September 1, 1987

Few organizations can use their basic, customer-serving network to develop new and untested communications technologies. But the Massachusetts Institute of Technology is no ordinary organization--and Project Athena is no ordinary network.

Athena, MIT's $70 million test of educational networking through linked high-resolution microcomputers, breaks ground in several ways. It is naturally supposed to be Olympian among intelligent networks, named, as it is, for Athena, the Greek goddess of wisdom.

It is also a move toward uniting the advantages of central processing and distributed processing. And Athena demonstrates how a world of high-bandwidth flexibility can ride on a larger, less intelligent network.

What made Project Athena intrinsically interesting from its announcement in May 1983 was its backing by both Digital Equipment Corp. (DEC) and IBM. Each contributed $25 million worth of equipment, and each is paying for five staff members. Both companies said then and repeat now that what fascinates them are new ways of teaching technology. Both firms are really most interested, however, in Athena's role as a test bed for networking.


From MIT's viewpoint, Athena's main purpose is educational. That boils down to networked pedagogical tools.

One example is Lingo, a computer language game resident on one of Athena's 30-odd applications servers, mainly DEC Microvax IIs.

To play Lingo you log on, call up the Lingo file, download it, and then watch a bit-mapped display of a room (your room in theory) be wrecked by a poltergeist. To put things straight, you type commands to the ghost. You type in German--that is what the ghostly software understands.


There are hundreds of educational and recreational programs like Lingo serving MIT undergraduates, and, according to the original plans for the project, there were supposed to be between 1,000 and 2,000 Athena microcomputers on the campus, says Project Director Steven R. Lerman. When MIT starts the 1988 academic year this September, there should be around 600, he says.

Also, development of Athena's software and file-server structure has moved somewhat more slowly than expected. So, while it was supposed to be a five-year test, Lerman says the full availability of Athena will not commence until this month.

That makes Project Athena a likely candidate for a multimillion-dollar extention request. Lerman agrees and says that MIT will make its formal decision on an extention sometime in the next few months. Interviews with DEC and IBM staffers indicate that their companies are likely to look favorably on such a request.

X for excellence

There's a good reason why. It was for a technical innovation that Athena has won its greatest kudos so far. They came in response to a powerful piece of linking software called the X Windowing System, X Windows for short.

Members of Project Athena's staff--mainly a DEC researcher on loan to MIT and an MIT scientist--developed X Windows to solve their network's biggest communications problem, which is a multitude of DEC Microvax IIs and IBM RT PC processors (and, subsequently, other vendors' 32-bit microcomputers).

Athena runs on MIT's campus network, mainly a hookup of Ethernets linked by a fiber optic backbone. Athena's basic networking software is Arpanet's TCP/IP (Transmission Control Protocol/Internetworking Protocol), which is also the basic software of MIT's campus network.

But Athena's processors didn't talk to each other in one important respect, although each of the processors uses the Berkeley Unix 4.2 operating system, which had previously been selected as the standard environment for the project.

The campus network works fine for passing messages. But for its primary role, which is teaching engineering, the Athena project had to pass graphics between different processors. Yet the project's staff realized that there was currently no recognized graphics standard that two different machines could use to communicate.

The tool that Athena devised, X Windows, has become the de facto standard for graphics exchange between high-resolution processors, says Ross Chapman, project manager for the user environments group at Apollo Computer.

''Unix is a great environment,'' he says, ''but it has no mechanics for bit-mapped displays. So every time you port a graphic, you have to put in a different code.''

Open window

X is more than a graphics standard; it is also a windowing standard similar to Microsoft Windows or to the desktop environment of an Apple Macintosh.

X Windows also allows a program to be sent from the user's microcomputer to a mainframe or supercomputer and then to return it in the form of a table, model, or formula to a user's window. While this processing is going on, the user can continue to work, tackling various tasks in other windows.

X Windows performs as if it were an ordinary piece of applications software, working with the operating system of each processor, Lerman says. The lower part of X Windows, called ''the server,'' is specific to the processor it serves and resides in that processor.

The upper portion of X Windows, ''the client,'' is a universal language of graphics protocols and operating instructions. It may reside either at the local site or in a central processor. It consists largely of a library of procedures.

As might be expected, this is a sizable piece of software. The client/library combination is approximately 2 Mbytes, Lerman says.

Now this is a fair amount of overhead, but it doesn't all have to reside in main memory. Portions of X Windows can be accessed from a local disk or from the network as needed, and the program's boon for storing the 2 Mbytes is significant.

Because of X Windows, a user wishing to speak to another type of computer need not go through protocol conversion. ''X Windows is the protocol,'' Lerman says.

Vendors quickly realized what the program could do for them. Among the firms now supporting it are Apollo, Dana Computers, Data General, DEC, Hewlett-Packard, MassComp, Stellar Computers, Sun Microsystems, Televideo Systems, and IBM. Service through servers. X Windows was a natural response to MIT's special networking dilemma. Another problem was how to provide information and software to a host of independent processors.

Most networks combine central information and local processing by hanging file servers on local area networks. Project Athena also uses file servers, but it cannot use them to serve single departments or single locations. MIT's campus network consists of multiple small LANs, but Athena must operate as one network, providing all functions across the campus.

Local file servers are thus only useful in remote locations such as off-campus housing, in this case, fraternity houses that access the network through leased 9.6-kbit/s analog circuits.

For other users, the information Athena dispenses comes from a linkup of more than 30 file servers (see figure), and with the exception of local print servers, each is accessible from any point on the network. In addition to the normal task of storing information--for instance, the educational software dispensed and data stored in personal file servers--many of the functions of the network are managed by servers, Lerman notes.

''For example, we have an authentication server--a server dedicated to controlling access to the network and to files,'' says Lerman. The need for central access control, he says, is because most microcomputers hooked to the network are open to any student who may wish to use them. ''And hence,'' he says, ''they cannot be trusted.''

For serving the servers, mainly for updating databases but also for restoring the network in case of failures, Athena has a service management server, a central server that, for example, tells the authentication server which students continue to be eligible to log on. Athena even has a server dedicated to locating software, electronic mail addresses, or other servers--and it can be used with simple, common-language questions.

The importance of Athena's servers, Lerman says, is that they bring back to distributed processing much of the control that was once possible only in hookups of terminals and central processors. ''If you had an infinitely fast central processor that could manage thousands of graphics displays, you could do what we are doing,'' he says.

''Personal computers are excellent for some things. They are great at graphics because they have a dedicated processor and provide relatively high amounts of power per user, but they are lousy at sharing. Central processors are just the reverse. We're trying to get the best of both worlds.''

No wonder DEC and IBM are interested in the project. ''Athena is essentially a distributed system,'' says George A. Champine, who heads DEC's Project Athena team. ''We find some of the ideas that have come from Athena very fruitful.''

IBM is particularly interested in Athena's combination of centralized and distributed processing through file servers, says Charles A. Salisbury, director of IBM's staff on the MIT campus, ''We find particularly important this connectivity between the two worlds.''

It seems that academic networks can be of more than mere academic importance--especially at MIT. As for Project Athena, it has had its problems, but the classical Greek name still stands for sagacity.

Diagram: Athena server structure


Copyright 1987 McGraw-Hill, Inc.