the GE switched services network

We currently find ourselves in something of a series, working our way from private lines to large private line systems like the four-wire private-line national warning system. Let's continue to build on the concept of the private line into large corporate systems.

In principle, a large organization in want of a private telephone system could build one out of a set of private lines and switches, such as under a Centrex CU (Customer Unit) arrangement. And this did happen: one common type of private line was the tie line, a private line used to link two switches together (which could be PABXs or Centrex) so that users could call from one to the other without using a conventional dial telephone line. This could save money, if usage of the tie line was heavy enough and especially if the two switches were far enough part that a standard call would be long-distance.

Consider a corporation with two large offices, each with a PABX. If they are in different local calling areas, calls between them placed by dial line would be long distance. If employees at the two offices call each other often, the long distance bills would add up to more than the fixed monthly cost of a tie line from one office to the other. There are a few different ways to solve this problem, such as getting WATS (wide area telephone service) at one or both offices, but it illustrates the general idea that getting a fixed private line can sometimes be a cost-saving measure compared to placing a lot of calls over standard dial service.

But what about a bigger organization, with many offices? You can imagine that getting a huge number of tie lines between different offices, planning where those tie lines should be located and how many were needed on each link, could become a feat of traffic engineering on par with the telephone company's own work. It might be easier to just pay the telephone company to work it out, and indeed, that's what large organizations often did.

So let's say the telephone company meets this request by designing a scheme of tie lines and Centrex exchanges. It's not so far off the mark to say that this describes AUTOVON. AUTOVON was a complete system of tandem exchanges and at least semi-private telephone lines provisioned by AT&T [1] for use by the military. The problem with this arrangement is that it is very expensive: the customer is paying the telephone companies to purchase, install, and maintain a huge amount of hardware, just for the customer's private use.

Now, comparing AUTOVON on a price basis is both difficult and unfair. Difficult because AUTOVON was paid for in a somewhat complex way, by the military paying a central, cumulative rate for the entire system and then performing cost recovery from individual user agencies and installations using a non-trivial cost assignment calculation. It is also often said that the Bell System themselves did not recover the full cost of AUTOVON from the military and that it was, to some extent, subsidized by other telephone services.

And I say unfair, because AUTOVON was more than just a private telephone network. It was a hardened private telephone network, with four-wire service and a precedence capability that required the development of novel equipment. It wasn't really expected to save money compared to the public telephone system, because it was acquired in order to provide capabilities that the public telephone system did not.

Still, we can safely say that AUTOVON was expensive. A 1979 study by the Defense Communications Agency, responsible for AUTOVON cost recovery at that time, comes out to an impressive total of $255,492,000 in AUTOVON operating costs for FY 1978. By way of example, the report puts the monthly service cost of a two-way capability from CONUS to Europe with priority precedence at $1,182. Obviously this example case is one of the most expensive, but I still shudder to imagine a monthly phone bill of over eleven hundred dollars in 1979 money. The military was willing to swing the stiff cost of AUTOVON because, first, it was the military and they were willing to swing the stiff cost of many things, and second because AUTOVON's military capabilities would be very expensive to build by any means. It was the Cold War, after all, and it could be said that outspending the Soviet Union was a military objective.

The situation was rather different when it came to non-military communications. The civilian federal government ran up some enormous telephone bills between its many offices, and initially considered purchasing an AUTOVON-like system to serve as a private network between federal offices. The concept simply wasn't cost effective, it likely would have increased the cost of federal telephone calls overall. The Federal Telecommunications System or FTS would eventually come to be, but not in the form of a private switched system. It is, after all, intuitive that cost savings would not come from installing a great deal of dedicated hardware. Rather, the Bell System would have to find a way to serve these large institutional customers with less investment. And that was the Common Control Switching Arrangement, or CCSA.

It is very tempting to draw an analogy between the CCSA and virtualization in contemporary computing, but it is probably more accurate to draw an analogy between the CCSA and LPARs in IBM hardware, or even more aptly, to virtualization's early precedent in the limited subdivision capability of Babbage's difference engine. Let's stop indulging the temptation and explain it more directly: a CCSA is created by configuring existing telephone switches to treat a subset of their lines as part of a separate network.

The technical details by which this was achieved varied significantly by the switch. CCSAs were introduced in the early 1960s and could be configured on the #5 crossbar exchange, where "configuration" consisted of strapping or jumpering certain components of the switch to operate independently of the others. CCSAs continued just about to the modern era, where configuration became a matter of selecting the appropriate lines in the business office system that generates configurations for computer-controlled exchanges.

I think that it's most interesting to examine the CCSA by way of example---by looking at a specific, real CCSA. BSP 310-200-007 I2 (1966) conveniently provides a directory of the code numbers that were used to identify CCSAs within the telephone system. Number 02 is FTS, the Federal Telecommunications System. I didn't bring it up without reason, the concept of the CCSA was developed in large part in order to bring the cost of FTS under control. We can ponder what happened to number 01, but I'm guessing that AT&T used that code for testing and thus reserved it, or maybe even to identify the public telephone system.

One could think of the normal, public telephone system as just another CCSA, although as I understand it this was not the nature of the actual implementation. Another appealing analogy for the CCSA is the VLAN, we could think of these CCSA network numbers as VLAN tags. In this analogy, the public telephone system is the PVLAN, sometimes called 0 or sometimes called 1 at the whim of vendors. If you are familiar with VLANs, that somewhat illuminates why I say that the public telephone network is not just another CCSA: it is the "untagged" network into which equipment not capable of CCSAs and lines not attached to a CCSA are presumed to exist. Anyway, that's all besides the point, what other CCSAs existed?

04, General Electric. 05, New York Central (railroad), 06 Lockheed, 07 State of California, 08 AUTOVON (used to facilitate expansions of AUTOVON over non-AUTOVON telephone infrastructure, as a more cost effective way to provide AUTOVON lines at smaller installations), 09 American Airlines, 10 Boeing, 11 Westinghouse, 12 Western Electric, 13 IBM, 14 North American Aviation. That's the complete list as of 1966, and while short, it is a who's-who of the industrial giants of the post-war United States. Plus the State of California. Most state governments used large Centrex-and-WATS arrangements, but some combination of the large size of California and GTE's different approach to the network steered them in the CCSA direction.

Of these CCSAs, I will focus on General Electric. There are two reasons: first, GE had an early and large CCSA---the largest CCSA outside of the federal government, at the time. Second, I was an intern at a failing GE business in a large, half-abandoned corporate campus [2] during the summer that the last vestige of the GE Switched Services Network, as AT&T called it, was retired. Among my scattershot duties was working on the decommissioning of the campus's Nortel PABX in favor of Cisco UCM. GE SSN would go with it, replaced by IP trunking between UCM sites.

In 1963, the GE SSN spanned fifteen central offices ranging from New York to Los Angeles, all #5 crossbars. It was intended to provide voice as well as data at 1200bps. Unlike some (mostly federal) CCSAs, it was designed to provide standard two-wire dial service only, without station-to-station four-wire connections or call precedence. In other words, it was a standard telephone network, but intended to make calls between GE offices more reliable and less costly than calls over the long-distance telephone network.

One of the complications of the GE SSN, and of CCSAs in general, is the diversity of telephone equipment in use across the different corporate offices. The GE of 1963 had Centrex service, step-by-step PABXs, crossbar PABXs, key systems, and manually operated PBXs. All of these were integrated into a 7-digit dialing scheme for GE SSN, with the NNX prefix (different from NXX used in the public telephone network by prohibiting a 0 or 1 in the second digit position) identifying a location on the network such as a PBX, and the four-digit subscriber number generally being the telephone's local extension, padded with arbitrary digits as needed to be four digits long. Of course, the details were less tidy, with smaller locations sharing prefixes and some locations acting almost like toll stations with single telephones on the GE SSN and selecting it by key.

In general, though, at PABX-served locations, extension users had a choice as to how to place their call. The configuration wasn't the same at every office, but the recommended practice was to use a 9 prefix (or "exit code") to dial on the public telephone network, and an 8 prefix to dial on the GE SSN. Most PABXs have some version of this capability: specific trunks can be selected for outgoing calls based on the dialing prefix.

At locations with manual exchanges and locations without compatible PABXs, GE SSN calls had to be placed with the assistance of the local PBX operator. Still other locations used a small PABX connected via tie line to a larger PABX at a larger office, in this case the dialing prefix "18" was recommended to first dial a trunk from the satellite PABX to the main PABX, and second from the main PABX to the telephone exchange providing GE SSN service.

Indeed, let's reflect a bit on the wiring scheme involved.

CCSAs were served by Offices, like the fifteen I mentioned for GE. FTS and AUTOVON had more offices (AUTOVON's CCSA office list tellingly includes the proper AUTOVON exchanges), but most CCSAs had fewer, sometimes only a handful. Between these offices, trunk capacity could be shared with normal telephone traffic, giving CCSAs a significant cost advantage. Individual phones (or PABXs or etc) on CCSAs needed to be connected to an actual CCSA office, though, in order to have access to the CCSA at all. This made practical CCSAs sort of a hybrid situation.

A given GE office might have lines running directly to the serving office, for example in New York City where there were two offices on the GE SSN. Offices that weren't near a 5XB included in the scheme, though, would need to somehow be connected up to one. The BSPs are not replete with details, but presumably this was done using the fairly conventional foreign exchange service.

I think I have mentioned this before but I will provide a very short summary. In the VoIP industry, it is extremely common to identify the "ends" of a telephone subscriber loop using the terms FXO and FXS, for Foreign Exchange Office and Foreign Exchange Station. Confusingly, the terms refer to where a given connection "goes" rather than where it "is," with the result that a Foreign Exchange Office connection is what you would plug into a phone, and a Foreign Exchange Station connection is what you would plug into a device like an ATA---something that provides talk battery, ringing, etc., the traditional role of the telephone office.

So the "Office" and "Station" part of those terms makes sense, besides the fact that they are arguably the opposite way around from what you would first think. But what about the Foreign Exchange? Well, these terms predate VoIP by decades, and were originally used to identify the ends of a foreign exchange service.

Foreign exchange was a specific type of private line that allowed a phone to be connected to a different central office from the one that physically served it. There are different reasons that this was useful, but a common one had to do with long-distance rates and suburban areas: it was historically common in large metro areas that suburbs could call into the city at local rates, but suburbs could not call into other suburbs at local rate... that would be a long distance call. You can see that this provides a bit of an economic advantage to city phones. So if you are, say, a plumber with your shop located in a suburb, you might pine after a big-city telephone line that would allow more of your prospective customers to call you for free. Foreign exchange could solve that problem.

When you ordered foreign exchange service, the telco connected your phone line, at the distribution frame of your local central office, directly to a private line. The private line went to a central office in the Big City, where it was connected at the distribution frame to a local line served by the switch. In practice there were complications and details to how this was set up, but this description gives you the idea: your telephone was now connected to the switch in a different central office from the one your local loop was actually connected to.

Foreign exchange service was expensive, because it took up private line capacity, so various combinations of WATS, InWATS, zenith numbers, toll-free numbers, etc. have pretty much replaced it. Some of the terminology got stuck in our modern telephone parlance, though. FXO and FXS were designations used by the telco to keep track of which ends of the private line needed to connect to what equipment. Why was I talking about this, though? Oh, right, because large CCSAs in practice also relied on what was basically foreign exchange service in order to connect outlying locations to CCSA offices.

It's a good thing, here, that most of these GE offices had PABXs. This limited the number of outside lines that needed to actually go to a CCSA office. What I am calling foreign exchange lines could also be viewed as tie lines, not really serving phones but providing trunks from PABXs to CCSA offices.

The nature of the CCSA is pretty much in its name: it is a switched private line service, but it makes use of common control equipment to minimize costs. I have tried to make terminology a little simpler here, but I have kept saying "GE SSN." Switched Services Network, or SSN, seems to have been a term used by the Bell System to refer to any private switched network. A CCSA was one of the ways of implementing an SSN, and seems to have been the most common throughout telephone history. There were not many truly private switched systems. AUTOVON could be considered an example, although it had requirements above and beyond typical telephone service. That leaves the FAA as a purer example, as the FAA used a significant number of private line services for both switched and unswitched communications between air traffic control sites and equipment.

Incidentally, a precursor to AUTOVON was called SCAN, the Switched Circuit Automatic Network. SCAN was a US Army four-wire system, because four-wire service was required for the cryptographic equipment of the era to function. AUTOVON seems to have inherited its four-wire nature directly from SCAN. Skimming through a telephone tariff almost always turns up some interesting details, one of them being that a few state telephone tariffs describe Switched Service Networks as being private line service based on either CCSA or SCAN. Given that these same 2024 telephone tariffs define SCAN as a federal government service for secure communications, this definition (and the presence of an entry for SCAN at all) seems to be purely a holdover from some fifty-year-old tariff documents. It does go to show that non-common control switched services networks were uncommon enough that telcos viewed AUTOVON as the odd exception to the CCSA rule.

I am trying not to get too tied up in the history of AUTOVON, because it is easily its whole own article. I do think it is fair to say that the CCSA emerged largely as a response to the high price of AUTOVON, as building FTS based on the pattern of AUTOVON was deemed completely unrealistic on a cost basis. FTS launched in 1963, not far behind AUTOVON at all, but consisted of CCSA service for long-distance calling with PABXs in government offices furnished under lease agreements. Once the heavy lifting had been done for FTS, it was natural to extend CCSA as an offering to large private companies.

FTS also holds some of the seeds of the Bell System's undoing. High costs and lackluster service plagued FTS in its early years. During the 1970s the General Service Administration, which was responsible for FTS, decided to introduce a competitive bidding process for long-distance capacity. That lead to companies like Western Union and MCI joining the network, and the introduction of least-cost routing to select the carrier. These ideas would also spread from government to private industry, helping to set up the industry-wide tensions that culminated in the 1982 breakup of AT&T.

A common term in the telephone industry is "Universal Service." In the modern world, universal service is understood to be the goal of providing telephone service to all customers. The Universal Service Fund, for example, levies a fee on telephone lines to subsidize the provision of telephone service to those who would otherwise be unable to afford it. This is a recent invention. 1960s documentation on CCSAs makes repeated reference to Universal Service, a very different form of the concept championed by AT&T more or less until its breakup: that everyone would be served by one, unified telephone system, the Bell System. During the early days of the telephone, before AT&T's monopoly was cemented, Universal Service was a rallying cry against competing telephone companies, whose independent networks interfered with the ability of any telephone to call any other. In the mid-century, it became a term somewhat like public switched telephone network (PSTN). CCSAs were capable of Universal Service, where desired, on a somewhat limited basis, in that CCSA exchanges could be configured to allow calling outside of the CCSA, into the public system.

It is amusing, then, that AT&T was so willing to abandon the ideal of universal service when their customers offered to pay for it. But that's business, and in the mid-century, AT&T was one of the biggest businesses in the world. CCSAs show us some of the ups and downs of the age of the telephone monopoly: CCSAs were an innovative concept that was rapidly developed and delivered, first to the government and then to private customers, over a span of just a few years. They were also frightfully expensive, and offered a new level of lock-in that kept customers from using competitive carriers.

It's hard to find any contemporary information about GE's private telephone system. It has almost entirely vanished from history, except in the form documented by the early BSPs. I don't even quite remember what it was called when I was at GE, I think it might have had "star" in the name. There was a printed directory to help you figure out the correct office code and means of transforming an extension in order to dial over the system. I don't think the copy I saw was recent, I'm not sure if a recent copy even existed. The system was barely used at all. It was replaced by a common modern arrangement, least cost routing with IP trunks.

When calling another GE office, the Cisco Call Manager installation would connect the call over the data network to the other office's Call Manager. Very practical, very easy to use, kind of boring.

[1] Even prior to divestiture, the practical construction and operation of these telephone systems was split between AT&T Long Lines and the telephone operating companies. Many, but not all, of these were subsidiaries of AT&T. In the case of AUTOVON, for example, we must consider that non-AT&T subsidiary GTE built and operated part of the system. I'm trying not to get bogged down in this complexity, but I'm also trying not to keep writing "AT&T" when referring to work done by multiple companies, some of them independent. Please do me the kindness of understanding that when I use terms like "the telephone company" or even "the Bell system" I am trying to encompass all of the parties, AT&T, Bell Operating Companies, independents, etc. that were involved in this work. The term I use may not be exactly correct.

[2] It was GE Intelligent Platforms, and the beautiful but poorly maintained corporate estate in Charlottesville, itself the remains of a failed joint venture, might have been a more suitable exterior for Severance than Bell Labs Holmdel.