Athena at MITanalysis
By Simson L. Garfinkel
Project Athena is a five-year experiment. As a result, major changes have been made in mid-stride. This has caused the system to evolve differently from original predictions and promises.
Many people feel the promises MIT made when Athena was first announced in May 1983 were unrealistic. Athena was to increase by more than an order of magnitude the amount of computer power at MIT. The project would give the Institute three thousand more terminals.
The networking would be so transparent that the system could be viewed as one large computer, with thousands of users working simultaneously. Alternatively, the final system could be viewed as thousands of personal computers, all sharing common disks, printers and other resources.
From the beginning, "coherence" was a key goal of Athena. Programs written on the IBM equipment had to run on the DEC equipment, and vice versa. The final system of three thousand terminals and workstations was to consist of both DEC and IBM equipment. "Coherence" meant that the entire system had to run together without any problems.
Today, almost two years into the project, Athena has gone a long way toward accomplishing many of its goals.
Five Microvax I's, 44 VAX 11/750s, and one VAX 11/780 serve 243 operational terminals at nine clusters. Many problems not anticipated have been uncovered and solved.
Although Athena has accomplished remarkable feats, it has fallen behind its original schedule. The current system bears only partial resemblance to its 1983 forecast.
Athena announced in the summer of 1983 that by January 1985 between five and eight IBM clusters consisting of over 500 IBM Personal Computers would be available for campus use. Athena also planned to give accounts to every undergraduate by that time.
Part of the reason behind the equipment shortfall lies with one of Athena's two major suppliers: IBM. According to sources both inside and outside of the project, IBM has often failed to meet deadlines.
The IBM PC/XT was to be a shell within which the IBM Athena workstation was to be built. High-speed co-processor cards and high-resolution graphics boards were to be piggybacked onto the IBM PCs.
These plans fell through because of problems with the co-processor card. An idea was then proposed to network the IBM PC/XTs with IBM 4341 mainframes. The 4341s would have acted as file servers and network managers. IBM's role in Athena had to be rethought, over the summer of 1984, when it was decided that Athena would not be receiving the IBM 4341 mainframes the project had intended to use.
The IBM 4341 mainframes were cancelled for a variety of reasons. Foremost, Athena was not able to support the machines. Supporting these machines requires a specially trained group to operate and maintain the computers' hardware and software. Athena was already severely understaffed.
Additionally, single 4341 mainframes would have required more floor space than Athena had available for entire clusters.
But officially, Athena cited technical problems as reasons for scrapping the 4341s.
The Use of PCs
Although Athena is using 63 IBM PC/XTs in the clusters, the XTs are running terminal-emulation programs that make them functionally equivalent to Heathkit H19 terminals. The Heathkit terminals cost almost a fifth of the price. While students can run private software on the IBM PC/XTs, not many do. There is no official program to encourage this use.
There have been marginal benefits to using IBM PC/XTs as terminals. They are more rugged than most terminals. In addition, students may use the PCs floppy disk drive to augment their file storage limit by storing data on removable floppy disks.
This ability is important in light of the restrictive limits on file space in the new Student Center cluster [see Jan. 23, 1985 entry in accompanying Athena calendar]. Currently, two floppy disks will hold more information than the standard Student Center account.
The remaining PC/XTs Athena received are used as independent computers, and are not in place in the clusters. Some laboratories use the PCs as data acquisition devices. Other PCs are being used to run special stand-alone programs such as drafting software in the Department of Architecture.
The ultimate disposition of the project's 150 IBM PC/XTs has yet to be determined.
While Athena will likely meet its latest deadline to offer accounts to every undergraduate in the Student Center cluster, nobody knows what computing environment will result.
Most feel that while the Athena VAXs in the Student Center will be slow due to the high user load, they will still be faster than MIT's Honeywell MULTICS, the computer on which students have traditionally received free accounts for text processing.
The Athena staff has had a number of major triumphs in its two years of existence. The list includes:
O+ X: The development of a networked window system (named X). "X has been developed from essentially nothing," Douglas J. Wilson '78 said. Robert W. Scheifler '76, a staff researcher at the Laboratory for Computer Science, brought X into its final form. DEC has done additional work on the project and may market the program in the near future.
O+ UNIX: The debugging of the majority of the UNIX operating system. The source code supplied with the Berkeley Standard Distribution 4.2 UNIX operating system did not match the object code files which accompanied them. When the source files were compiled, they did not run properly. David G. Grubbs '76, Athena Systems Programmer, spent a considerable amount of time getting the operating system to compile and operate.
O+ Reliability: The development of one of the most stable UNIXs in existence. Athena machines rarely crash, and typically run until shutdown for maintenance.
O+ Third party software: The debugging of a large body of third party software. Although Project Athena has purchased and been given a good amount of third-party software, much of it has had to be debugged to get it to function properly.
O+ Hardware installation: Installations of nine terminal clusters and six machine rooms across campus. Athena has installed nine terminal clusters across the MIT campus, with almost 250 terminals total. Along with Information Services, it has installed a campus fiber-optic network which extends from 545 Tech Square to Building E40, with connections to Buildings 4, 11, 38, and the Student Center.
According to Doug Wilson, Athena is expecting 2700 users this term. Last Thursday, Athena registered 50 students an hour for course accounts.
Codex Inc. has joined IBM and DEC as a primary Athena hardware supplier, according to Wilson. Codex will design, with Athena staff, a new generation of network "gateways" for use on the campus network. Until the Codex gateways are ready, Athena plans to use Microvaxes and PDP 11/23s as necessary.
Looking to the future
Athena is about to release a remote procedure call (RPC) compiler. This software package would allow programs running on one computer to call subroutines running on another computer. The release, which will include a complete programmer's manual, is scheduled for later this semester, according to Wilson.
The RPC compiler will allow programmers to "build distributed applications without knowing anything about networking," Wilson said.
Athena also plans to release INFORMIX, a new database system.
The project awaits the delivery of 80 IBM PC/ATs. These machines will run XENIX, a version of the UNIX operating system. Larry W. Allen, a staff researcher at LCS, recently wrote programs to allow the PC/ATs to communicate over the Athena network. IBM's Yorktown Heights research laboratory modified the programs at Athena's request.
The PC/ATs will support TCP/IP protocol as well as TELNET and TFTP (Trivial File Transfer Program) programs. The TELNET program allows a PC/AT user to log onto any other Athena host, while the TFTP enables file transfer between the two machines, according to Wilson.
There could be a major change this fall in the way student accounts are distributed, Wilson said. Clusters in the Student Center, and Buildings 1 and 66 will most likely be dedicated to student use.
Course material will reside on special computers in these clusters, rather than in special clusters for courses as is now the case. Students will no longer own an account for each course they are taking. Instead, they will have one account with access to all course material they desire.
A conflict between the need to build a stable campus computing resource and the desire to experiment in the use of computers in education lies at the heart of Athena. Which is the real purpose of Athena?
"It's both," Wilson said. "The permanent campus resource is the network. The computer infrastructure is the network and the network services." Wilson hastened to point out the large amount of course software which has already been written for the Athena system.
The Athena system is in a state of flux. Over the next few semesters, the emphasis will shift away from time-shared minicomputers to single-user workstations. The course Athena will take is still unknown.