relationship with the FAA, the accomplishment of this objective can be performed in a period we estimate to be within 2 years. For your cost estimation questions, as I noted earlier today, I have a simple formula. Each machine would cost roughly about $4 million. If you multiply that, times the number of sites you would like to see in the interim period, times 2 for redundancy, for the obvious fail-safe factor, plus the software expense of rehosting NAS software, which we think to be less than half a million dollars, you can come up with a pretty rough estimate of what the cost would be for an interim system.

Mr. GLICKMAN. What is it?

Mr. MARTINCIC. I don't know how many centers you would like to do in it. For instance, if you have five capacity centers-that is centers now, that are running over capacity, if you want to do it in five, that is the formula you would use.

Mr. GLICKMAN. What is the aggregate amount per center is what I am trying to get at.

Mr. MARTINCIC. $8 million plus an initial investment of about a half a million dollars to convert the NAS software.

Mr. GLICKMAN. And how many total centers do we have, 21? So you are talking about 100 and-anywhere between $160 and $200 million to do it with all of the centers?

Mr. MARTINCIC. That is correct.

Mr. GLICKMAN. OK.

Mr. MARTINCIC. Rather than address the lessons from other Government procurements as outlined by the subcommittee, I want to discuss a bit, the lessons that you should learn from the last NAS procurement from IBM. The NAS system developed in the mid1960's was not designed with the ability to withstand an infinite number of functional additions. Many of today's software problems are the result of constant code modification to facilitate additional function.

And, in conjunction with that, item D, the application software; that is, the components of software responsible for various functions, must be written and maintained in a level permitting portability, modularity, and maintainability. The current software, at one time, was written in a higher level language. It was called JOVIAL. It was then translated to a machine level code. That was part of resource buyback. Done in the hopes of efficiency, it really affected portability, modularity and maintainability.

We think that industry, rather than government, should represent the final hardware/software design as a functional solution to the requirements of the procurement. Government should insure that the industry approach permits growth through system component modularity. No component replacement should require redesign of all the other components.

Standard, off-the-shelf, proven hardware and proven software techniques should be the rule in future acquisitions. The decision to use unique, special design instructions outside the repertoire of the standard 370 instruction set and the special design of the hardware component interface level has made upgrade of the system all but impossible, even within the incumbent vendor's line of equipment.

In summary, the FAA, including all of its member organizations of controllers, technicians, and management, deserve an enormous amount of credit for its record of preserving air safety.

The complexity of the NAS system is such that few organizations and especially few people can be described as expert.

Neither Amdahl nor we as their employees, can offer expert advice or criticism in areas of air safety. Our reputation is based instead in the application of current technology to large-scale computer systems and the accompanying hardware and software support.

The Amdahl Corp. views the modernization plans of the FAA as we would most critical mission, computer-dependent applications. Admittedly, the life-dependent nature of the FAA's mission creates special pressures on those who would support that mission. Amdahl considers any revolutionary solution to the problem of modernizing the ATC computer system to be a plan of unmanageable proportion. Only an evolutionary approach can hope to be successful, especially in a transitional process.

It is the opinion of our organization that total redesign and replacement of the NAS system is prudent only by first rehosting the current NAS software on modern hardware. Our estimate of present value of the Federal investment in NAS software is near a half billion dollars. Further, current computers, in addition to of fering much great reliability and maintainability, offer power and performance that would permit simultaneous support of the current NAS software and a newly developing or developed hard or software.

Although FAA view efforts to rehost NAS as prolonging the life of a product that requires replacement, our belief is that the benefits to FAA include: (a) A vehicle of transition to the replacement system; (b) a preservation of a substantial investment; (c) an option of added capacity in centers of high utilization; and (d) a hands-on experience with new technology.

And implementation of future functions will no longer be contingent on software improvement programs or any buy-back capacity as stated by FAA.

[Complete statement follows:]

Mr. GLICKMAN. Thank you. So that I clearly understand where you are coming from, do you believe that over the long term, there does need to be a generally complete replacement?

Mr. MARTINCIC. Yes; I do.

Mr. GLICKMAN. You seem to opt in short term from kind of a modified interim replacement not of the scope mentioned by the GAO, that is, basically taking the 9020's and modifying them. Mr. MARTINCIC. No.

Mr. GLICKMAN. What do you propose then?

Mr. MARTINCIC. I think everything in the center, from the terminal that the flight stips come out of, all the way to the broadband radar that feeds the 9020 should be replaced, both because it is obsolete equipment and because the software has also been patched to the point that I don't think it can be maintained much longer. But in our opinion-I'm looking just at the 9020 computing complex-that portion of the hardware that does the computation for the control can be replaced, possibly within 18 months, using the

standard 370 instruction set that is used by the 9020. There are many other items of hardware in that center that should be replaced, either simultaneous or in some kind of phased pattern. Mr. GLICKMAN. Well, you have to excuse my ignorance. Willunder your proposal, will the 9020's themselves be replaced?

Mr. MARTINCIC. Yes; they would leave, or they could leave the

room.

Mr. GLICKMAN. And do you think that this could be done for a figure under $200 million?

Mr. MARTINCIC. Absolutely.

Mr. GLICKMAN. And how long would this protect us for?

Mr. MARTINCIC. This interim buy could, in fact, be the final computer solution to the 9020 replacement system. The family of computers that we see now in the marketplace, including our own-and by the way, we feel that there are only three manufacturers that can really offer this, have just been announced within the last 6 months. Certainly, this new family will be good for 20 years.

Mr. GLICKMAN. Who are those three?

Mr. MARTINCIC. Amdahl Corp., IBM Corp., National Semiconductor. They all use the standard 370 instruction set, which is a basis for that entire software package that runs in FAA.

Mr. GLICKMAN. Do you believe that with the existing 9020's, that it is possible to keep them fixable, operable, the next 5, 6, 7, 8 years; or do you think that is not possible?

Mr. MARTINCIC. I know that there are problems getting parts for the 9020's. The line of equipment that the 9020 is based on-you cannot get memory for it, for instance, anymore. I don't know about other parts, of that manufacturer's line.

I really don't know, but if it has to be replaced quickly, it can't be replaced any shorter than about 18 months to 2 years, so if the requirement ever does come up, I think it is in the best interest of FAA to be be able to have an option to operate.

Mr. GLICKMAN. OK. But to go back to my other question I asked, is it possible to operate the 9020's safely for the next 5, 6, or 7 years, without making major modifications?

Mr. MARTINCIC. I don't know. Our approch to this, as a problem, has been as a computer problem. The amount of knowledge we have on the 9020 is more from the engineering and technical viewpoint, than on the capacity of the 9020 and reliability.

Mr. GLICKMAN. From your knowledge of computer systems, would you say that the 9020's are experiencing an unusually high outage rate?

Mr. MARTINCIC. Not for the family of equipment that they represent. The technology that the 360, 50, and 65, which are the component bases of the 9020-I don't think they are experiencing a failure rate any larger than those of that family of equipment. However, the amount of availability that's in the 9020 is somewhere around 99 percent, as we can see the figures; and admittedly, we look at three different reports, and there are three different figures. But that kind of reliability is available today, in a single system, where the requirement of the 9020 is that you have five or six computers ganged together such that if one fails, the other can kick in-that kind of approach.

Today's availability of systems is in the neighborhood of 99.6 or 99.7 percent for many manufacturers.

Mr. GLICKMAN. Mr. Dunn?

Mr. DUNN. In talking about reliability-and you are approaching this strictly from a computer standpoint-several of the other witnesses have talked about safety. You use the words: "currently, ATC reliability is quite good by today's standards."

I assume by "today's standards" you're talking about computer standards?

Mr. MARTINCIC. That is correct.

Mr. DUNN. You're not talking about the number of lives that we might or might not have lost?

Mr. MARTINCIC. That is correct.

Mr. DUNN. No other questions.

Mr. GLICKMAN. You mentioned this briefly. Just one last question. How many activity-increasing improvements like the AERA could be installed on the interim computer you envisioned?

Mr. MARTINCIC. Well, the computer of today represents 20 to 30 times the power of one of the 360 components of the 9020 system. Actually, someone that would replace a 9020 would replace seven or eight boxes in one replacement. That is to say, there may be four computer elements and three I/O computer elements which comprise the 9020 system.

A replacement system would offer 5 to 10 to 15 times that power. That is available today. With 5 to 10 to 15 times that power, we think that most of the things that the FAA would like to install could probably be done.

But more importantly, all of the development of those packages could be done simultaneous to handling air traffic control.

Mr. GLICKMAN. OK. Thank you very much for your testimony. Has Dr. Goldmuntz arrived yet in the room?

[No response.]

Mr. GLICKMAN. Well, why don't we recess for 5 minutes. We'll wait until he comes.

[Recess.]

Mr. DUNN [presiding]. In the absence of our chairman, we will go ahead.

Dr. Goldmuntz, your testimony may be submitted in full, or read, whichever you think is best.

Dr. GOLDMUNTZ. Thank you very much.

Mr. Chairman, it is a pleasure to appear before you. I would just as soon extrapolate my testimony, and file it as it is, with the committee.

Mr. DUNN. All right.

STATEMENT OF DR. LAWRENCE GOLDMUNTZ, PRESIDENT,

ECONOMICS & SCIENCE, INC.

Dr. GOLDMUNTZ. My name is Lawrence Goldmuntz. I am president of Economics & Science Planning, a Washington-based consulting firm that has been concerned, among other matters, with air traffic control issues for almost a decade.

Mr. GLICKMAN [presiding]. Go ahead. I'm sorry I'm late.

Dr. GOLDMUNTZ. I was just going to say, it's a pleasure to appear before you.

We coordinated the recent FAA-sponsored research for a consensus of users' views on new engineering and development initiatives, and we led the team that prepared the Automatic En Route Air Traffic Control-AERA-concept document, which I understand has been published by FAA under the numerals FAA-EM-81-3; it should be available shortly, if it isn't available at the present time. We are now contributing to the development of an AERA program plan and its relationship to the computer modernization program.

I really have very few words to say, other than the written testimony, and I would like to say them, then make myself available for questions.

Mr. GLICKMAN. Surely. Your whole statement, as Mr. Dunn said, is in the record, so you need not read the whole thing.

Dr. GOLDMUNTZ. The thrust of my testimony is that it would be a shame to replicate the current system in the computer replacement program. It might make it slightly more reliable as far as the computer complex is concerned, or make it capable of handling more traffic; but it would do nothing to change the way the air traffic control system operates.

It would be a shame to do that, in view of the great progress-in my mind that has been made in making the air traffic control process more productive and safer; more productive for both the users and the operators of the system.

Based on the work FAA has already done-ATARS, en route metering, et cetera-the AERA concept team felt that it was possible, within the state of the art, current computers-DABS, datalink, et cetera-to have an air traffic control system that would respond to the wishes of the users, with respect to flight profile and direct routing and thereby save an enormous amount of money for them.

Our estimate-and it is only an estimate-is that the users, by being permitted to fly the flight trajectories and routes that they wish, could save on the order of 3 percent of their annual fuel bill.

The fuel bill for the commercial airlines in 1980 was $10 billion. Therefore they could save approximately $300 million per year. If you look at the present value of that fuel saving, due to more efficient routing, it turns out to be approximately $3 billion-just from that this single increase in productivity to be gained by more efficient routing.

Mr. DUNN. Can I interrupt you?

Dr. GOLDMUNTZ. By all means.

Mr. DUNN. You're talking about AERA and optimum routing. Can you give me an example of exactly how that would work, in terms of what is the optimum fuel-efficient routing into Washington National?

I'm not sure what AERA could do that the present computer

Dr. GOLDMUNTZ. Let me give you one example. Up until very recently, people who were flying from New York to Washington would be restricted to altitudes of 16,000 or 17,000 feet. The reason they were restricted to those flight levels was because of other aircraft above them-at 24,000 or 26,000 feet.

That has been somewhat changed because of better coordination procedures that have been recently invoked. This traffic can now