On Distributed Communications Series
VIII. The Multiplexing Station
VI. Major Equipment Complement
This list provides a brief summary of the components implied by the first rough-cut estimate of the proposed system. It is not meant to be complete, as not all the circuits have been examined in equivalent depth. It is a "once-over-lightly" examination of the amount of equipment that each unit performing some function that needs to be performed will probably take. However, in the past this method of analyzing digital equipment has been found surprisingly accurate--usually within a fractional order of magnitude. The number of components required to perform the necessary functions is, upon a second look, usually thought to have been underestimated by a large factor. But, later, when the detailed design of the system is completed by a competent, and ingenious logical designer able to take advantage of minimization of logical functions, a saving roughly equivalent to the underestimate seems to occur.
All that can be said at this point with any confidence is we anticipate that a unit occupying about 150 cu ft, and costing about $300,000 exclusive of research and development, would be on the approximate order of magnitude of equipment needed to build Multiplexing Stations that permit the interconnection of 1024 simultaneous subscribers into the system.
This low price can only come about after the equipment has been fully developed, tested, and production quantities manufactured. In order to build such low-cost equipment, it will also be necessary to use industrial-grade components in lieu of standard high-priced military-approved components. It is felt to be much cheaper to maintain a good atmospheric environment for electronic equipment than it is to tighten the specification of the environment and, in turn, the components used.
Again, it should be emphasized that no detailed complete design has been made; this is only a first-cut estimate. Only those processing operations that are performed continually and quickly, and which require special units, have been examined. All the operations performed in setting up a call appear to be within the constraints of the Central Processor, Unit J, which is almost a general-purpose computer. There appears to be sufficient time available for performing the extremely complex operations. We have attempted to limit our examinations only to those operations that are highly repetitive and which set the boundaries on the amount of equipment required.
Insufficiently Examined Areas
There are certain areas in particular that merit more detailed examination in order to better specify the actual number of components required. These embrace those functions difficult to analyze without more labor than would be appropriate at this time. They include:
- Exactly how much core memory is required in the Central Processor? Some of this memory is needed to perform the signaling functions and special error tracing operations; most will be set aside for handling peak loads of outgoing Message Blocks. While we intuitively feel that about 2000 32-bit words will be sufficient (and will avoid the necessity of dropping messages due to instantaneous overloads), more examination is indicated.
- Message Blocks from 32 separate lines are interlaced on a single drum band, so the bits forming a single Message Block will not appear in sequence on the drum. However, it is anticipated that reciprocal operations at the transmitting and receiving Multiplexing Stations will eliminate any problems originating from this cause. But, if such problems are encountered, there are several possible ways out, but these have only been briefly touched upon. No single best method is being suggested at this time because of a limited amount of time for analysis of the alternatives.
- We are not wholly satisfied with the expected reliability of the Multiplexing Station in its present configuration. For example, in the event of major failure, the number of usable lines drops to a very small number, and part of the secrecy protection is lost until repairs are made. The use of acoustical delay line registers, instead of one big drum, plus the replicating of more of the elements of the Multiplexing Station, would permit more satisfactory operation in the event of failures. A failure mode analysis should be one of the first tasks undertaken.
- We would like to see low-cost, cryptographically secure telephones built specifically to operate with this system, so that complete two-level security can be offered to even subscribers remote from the Multiplexing Station.
- We would like to see work directed toward lowering the cost of conventional hard-copy input devices, such as teletypewriters. All the savings and advantages of an all-digital system cannot be obtained unless the subscriber end-devices are designed with such systems initially in mind.
The individual component failure rates anticipated in the Multiplexing Station are essentially those of the components used in the manufacture of the Switching Node. We do, however, also have a magnetic drum with the servo-unit plus a small magnetic tape recorder that have a limited life--perhaps on the order of five years between major overhauls. Many of the failures we anticipate will occur in elements such as individual line units which will not take down an entire Multiplexing Station. However, until the equipment has been redesigned on the basis of types of failure, and the reduced capability modes of operation studied in detail, it may be misleading to suggest any meantime-to-failure estimates for the Multiplexing Station.
The best that can be done is to state that the complexity of this equipment is roughly about two or three times that of the Switching Node and could, at worst, be correspondingly less reliable. But, on the other hand, there are more inherent safeguards and standby provisions in this equipment. To be safe, however, we should plan; therefore, it would seem highly advisable to have the most important users tie into two or more Multiplexing Stations.