Xerox Finds Program Is TOPs for Private Network Savings.
The software processes actual call traffic data to produce minimum-cost, long-distance service configurations. What is unusual about it is that it does not use any statistical modeling. There are no assumptions, for example, as to how many calls were placed during the busiest hour of the day. The program takes each call and routes it in the same manner as the actual phone network. The computer program reads each call, routes it properly (taking into account all other active calls at that particular moment), and then records information about the calls and their costs. In theory, the program will completely mimic the actual call traffic. In practice, the program is generally accurate to within one percent.
When Xerox installed its present EPSCS (Enhanced Private Switched Communications System), which it calls Intelnet, in the late 1970s, EPSCS was a new type of AT&T network that allowed Xerox to capture nearly 100 percent of its call data for the first time. Since the network was inexpensive to run, only basic network reports were needed to assure that call blockage was minimized.
In the early 1980s, Xerox, as most large corporations, faced a multitude of cost increases on its network as a result of several factors: the demise of Telpak, WATS increases and drastic increases in private-line rates. Xerox's seven-node network grew to an $18 million yearly expense. The easy-to-run private network, which provided service to the corporation at substantial discount compared to the public network, came under tremendous economic pressure.
To offset these pressures, Xerox investigated carriers such as MCI and GTE Sprint. Satellite service was inauguratd with RCA and American Satellite. Sophisticated multiplexing gear helped reduce costs too, but these actions were just a start. Opportunities for Savings
One area where substantial savings opporunities were known to exist was in the routing of traffic from Xerox locations to Xerox customers (off-net traffic). In previous years, the economics of the network favored keeping a call on leased lines as long as possible and then entering the public network as close to the final destination as possible. For example, a call from Rochester, New York (a major Xerox location) to a customer in Fort Collins, Colorado, would first travel to Chicago and then to a WATS line in Chicago, where it would enter the public network. Over time, the economics changed and the cost to Xerox of carrying the call on its own network became greater than the cost of using WATS from Rochester (see Figure E).
Massive projects were undertaken to study reconfigurations of the network. The reports available required extensive further analysis done by hand. Although some positive changes came from these studies, it was clear that a computer program was required to obtain the needed accuracy and speed. The bulk of the savings simply could not be realized by hand.
Several options were considered. Xerox considered writing the software itself, but quickly realized the task was monumental. Next, a search was made for existing software. Although several fine packages were reviewed, Xerox chose the Teletraffic Optimizer Program (TOP) from The DMW Group in Ann Arbor, Michigan, because it used no statistical approximations at any time. Although TOP was relatively new to the marketplace, a reference call to another large corporation that was using TOP confirmed its accuracy.
Initially, Xerox management was reluctant to spend money on a non-traditional software package. Its bold step, however, paid off. TOP paid for itself in the first six months of operation and went on to deliver far more savings and valuable information than had been initially envisioned.
But TOP wasn't without its trials. TOP is a large program, consisting of pre-processing modules written in COBOL, dozens of data files and a Fortran simulator. One might think of Fortran as "antiquated," but we found the simulator to run very fast under the IBM VS operating system.
To be fair, it must be stated that installing the TOP package was tougher than either Xerox or DMW had predicted. The package had never before been used in an IBM environment. It took time to write the correct JCL (job control language) to perform the link-edits in the proper order. With time, the proper sequence was discovered. However, these problems caused the installation to run beyond the price Xerox had contracted for in the installation. Nevertheless, DMW continued working at its own expense until these problems were solved.
The first benefits from TOP came from the collection of information needed to create the data bases. Every line of the network had to be cataloged and priced. Several lines were found that had not carried any calls for months, and Xerox quickly cancelled them.
The first task assigned to TOP was to verify the proposed routing changes for off-net traffic. The simulator made this verification and, in fact, led to several changes that produced $5,000-per-month additional savings for the project. When the proposed routing changes had been completed, a check was made to determine whether what the simulator predicted had actually happened--it had. The consistent accuracy of TOP gave Xerox the confidence to try more sophisticated routing changes that could not be evaluated by hand.
Thus, second and third routing change projects followed. Savings on both of these additional projects were the same as or greater than on the first. The second project also had many side benefits.
First, the simulator showed the precise economics for overflowing from one WATS band to another. Along with this, errors in existing routing patterns were discovered and corrected.
Second, a long-standing quality problem was solved. Xerox had been using an ingenious method of allowing calls to traverse the Xerox network to terminate first in a PBX and then go on to be completed as local calls from that PBX. This practice is very economical, but sometimes the quality of the connections is poor. Using the simulator, ways were found to minimize the cost of rerouting this traffic to solve the quality problem.
Third, as more and more traffic was switched from expensive fixed circuitry to usage-sensitive circuitry, the infrastructure of the network had to be downsized (see Figure D). In re-sizing this structure, several arrangements were attempted. The simulator showed a new price for each rearrangement. By watching the cost result of each iteration, the point of diminishing returns was determined. Thus, the lowest possible cost was definitely found.
Another network situation where TOP was effective was evaluating the exact number of phone lines between every Xerox location and the national network. Figure G is a typical example of a TOP printout of a Xerox location. This location happened to have 10 access lines. Since access lines have no alternate-routing options, all traffic is listed in the "Choice I" column. Quick extrapolation will tell the user whether or not an eleventh line is needed. To simplify slightly, let us assume that 100 additional minutes of calling could be handled by an eleventh line. If that line cost $500 per month, then it could be valued at $5 per minute--whwch is rather high. But more than this, the cost-per-minute figure is useful in making comparisons among all the locations. Locations close to the network can be allowed additional lines more economically than those locations distant from the network.
Of course, this can be taken too far. Some locations could be reduced to so few lines as to interfere with normal business. Thus, certain minimum standards must also be considered.
One project that was already underway when the simulator was purchased was the addition of an eighth network-switching node in Atlanta. This project saved money and improve service at the same time (see Figure C). Altanta was an intuitive reconfiguration. The very long access lines from Miami to Washington, DC, and the great number of Xerox locations in and around Atlanta suggested Atlanta as a node. Nearby three person-months of hand study had already been invested in estimating a possible savings of $7,000 per month.
When the Atlanta situation was evaluated using the simulator, the savings figure was increased by $4,000 per month, and the cost of making Atlanta a node was reduced as well. Knowing that a substantial increase in telephone traffic was unlikely by the time the switching node was to be activated, Xerox did not have to install any spare circuitry beyond what was needed. It is common, when making major network changes, to install about 25 percent more circuitry than is expected to be necessary to cover uncertainties. Atlanta was brought into the network without the additional circuitry. The resulting traffic was exactly as predicted.
The Atlanta node was only one example of the power of the simulator to evaluate new configurations. Xerox used the simulator to determine whether each locations was attached to the proper switching node. The simulator was used to evaluate several proposals to replace all or part of Xerox's network with satellite circuitry and/or different switching nodes featuring the most advanced routing capabilities. None of these proposals was a problem for the simulator; it handled such features as queuing and multiple time-of-day changes, as well as complicated screening, without difficulty. Proved on Economical Answer
In sum, Xerox was exceptionally pleased with the results of its $40,000 investment in the TOP simulator.
As well it should be. As mentioned earlier, during the early 1980x Xerox was faced with a flurry of rapidly increasing network costs being brought on by the end of the ubiquitous Telpak service, coupled with dramatic increases in WATS and private-line rates.
With seven modes and a yearly expenditure of $18 million, the economic pressures mounted considerably.
As we saw, implementation of DMW Group's Teletraffic Optimizer Program slashed the expense by more than a million dollars, yet improved the quality of the already-excellent network serving Xerox's widely scattered locations.
|Printer friendly Cite/link Email Feedback|
|Date:||Oct 1, 1984|
|Previous Article:||New International Satellite Service Players Wait in Wings.|
|Next Article:||Deciding Whether Your Local-Area Network Should Be PBX-Based or CPU-Controlled?|