TPF Software is a Lifesaver
by Ronald V. Rose

On May 14th at 3:35 a.m., the New York City Police Department turned it's newly migrated TPF 3.1 (PTF 11) system over to its users. Just another hardware/software upgrade? Don't bet your life on it! This application was the New York City Emergency 911 system, called SPRINT (for Special Police Radio Inquiry NeTwork).

SPRINT doesn't handle a tremendous volume of messages compared to most TPF shops, but the calls are more urgent by far. Most are requests for help from the police or from hospitals by citizens in distress. When Bedford Associates arrived to perform the migration under contract to IBM, it was with an appreciation of the scope of the SPRINT function, and the potential impact of the conversion, in mind. Everybody was aware that problems could have tragic implications for the people of the city. We also realized we could end up with media coverage the likes of which most data processing professionals could only imagine. "Let's not make the cover of the New York Times," was a common theme among users and technical staff alike.

Because of the critical nature of the application, some interesting design approaches are incorporated into this TPF system. Each of these characteristics had to be thoroughly analyzed by the Bedford technical team. For example, to minimize the complexity of the application, and the underlying infrastructure that supports it, the concept of long-term pools is replaced by a process of rotating through fixed pools of records. That provides the advantage of eliminating the running of Recoup type utilities. "Well," you say, "where is the fun in that?" Perhaps the architecture is slightly less exciting than most TPF systems, but it is also fairly bullet-proof, if you'll excuse the expression. Total down-time, planned and unplanned, has been a mere 3/10ths of a percent! That's quite an achievement.

Bedford/IBM and the NYCPD had to migrate to a new hardware/systems software platform with minimal impact during the conversion itself, while maintaining maximum availability and reliability of the SPRINT application and system. We were successful in these goals for several reasons, perhaps the most important being thorough testing. "If you had to single out anything we did in particular that was right, it would be that we definitely tested the new system extensively and thoroughly," comments Tony Pece, the manager of the SPRINT area. This task involved the close coordination and support of the End-User community, lead by Claire Maroney. Albie Hohauser, who helped to coordinate the testing activities states, "We were lucky enough to get two excellent end-users dedicated to the project early on in the development cycle, and that really helped us a great deal."

After complete integration testing by the Bedford technical team and the 2 expert end-users, an extensive battery of 10 tests was performed with 30 users attached to the new system. All possible terminal types were attached, and each type of user conducted thorough testing to make sure the converted application was reliable. While these tests were going on, hardware fallback tests were run on the system to make sure there was no discernable impact on the operation of SPRINT.

The scheduling of the cutover involved some differences with traditional TPF systems as well. When is the best time to cut over a system? Did I hear someone say late at night over a weekend? The peak hours for transactions for NYCPD are late at night on Friday and Saturday. Naturally, steps were taken to minimize the actual outage, and the original TPF system was kept in reserve although it was never needed. Thirty additional end-users were asked to come in to facilitate the switch and provide extra support during the conversion process.

All the precautions taken in the testing and in the cutover process, served well to make this a smooth TPF conversion. Adds Bedford's President, Ed Gehrlein, "It just goes to show that good teamwork between end-users and technical people, and thorough testing procedures, insure that even challenging software projects, like this one, are successful."

To paraphrase the movies, "there are a million stories in the big city", and thanks to proper planning and testing this turned out NOT to be one of them.

Ronald Rose is a Project Manager for Bedford Associates, Inc. He was the Bedford Project Manager on the EVA Airlines ALCS project, and the Westin ALCS project, and has been involved in TPF/ALCS management for roughly 15 years.