RAID Protects Paging System For Telecommunications Company

A large telecommunications company faced some severe technical problems. A department of the company handled paging systems which transferred messages via the Internet. With this Internet-based pager system, using a browser, a user needed only to write a message, key into a PC the 10-digit number of a pager and press the "send" button. The message would appear on the receiver's pager. Alternatively, the user could use e-mail to send text. Unfortunately, the company used a legacy system that was load-sensitive. That meant that, as the message load increased, the system became less reliable. As a result, the system often went down twice a week on good weeks and 20 times a week on bad ones. And, these crashes were occurring despite the fact that the system was handling only a small number of messages - about 3,000 a day. Obviously, when the system crashed, messages could neither be sent nor received, and company technicians would have to repair the system each time. Clearly, the situation was intolerable.

Fortunately, the telecommunications company had done business for over two years with a systems integrator from Plano, Texas. Logical Approach, Inc., which built and installed a solution to this problem, specializes in technical software development, typically for engineering environments. Al Hopper is an engineering consultant for Logical Approach and the person primarily responsible for the new system, dubbed "Internet Gateway". He defined the problem which faced him: "Uptime is the name of the game for telecommunications companies." He needed to create a system which offered high availability and reliability on a 24 hour-per-day, seven day-per-week, 52 week-per-year basis.

Building A New System To Solve The Problem
Hopper explains that the legacy system was basically an older, slower, DOS-based system. Logical Approach replaced that system with multi-user, multi-tasking, UNIX-based workstations and servers. The systems integrator also added TCP/IP (Transmission Control Protocol/Internet Protocol), a way to internetwork dissimilar systems. At the same time, the client was building/deploying a nationwide frame relay network. (This is a high-speed, packet-switching protocol used in wide-area networks.) Basically, Logical Approach facilitated the migration from the older system to a fiber optic client/server network.

With regard to the components, Hopper's company originally designed Gateway with a dual processor Sun Enterprise II computer, then upgraded to a Sun Enterprise 4000. The E4000 provides many other services as well as running the Internet Paging Gateway. The E4000 has five processors and four GB of RAM plus 250 GB of storage. Attached to the Enterprise 4000 are five CRD-5500 RAID (Redundant Array of Independent Disks) controllers from CMD. The CRD-5500's "active/active" redundant controller option doubles the throughput of the RAID system and provides fault tolerance. The CRD-5500 uses RAID Level 5 to store Oracle data files.

What Was Challenging About The Installation?
Hopper says that there were two challenges to the installation. The first was the need of the system to be able to handle live traffic. Sometimes the e-mail which arrived was malconfigured, unreadable gibberish. The system had to be able to handle this e-mail 24 hours a day without crashing. The second challenge was protecting the paging infrastructure from "runaway mail" messages. For example, a large ISP (Internet Service Provider) automated a system to report when a Web server went down. One day, the ISP lost power for an extended period. Consequently, all 1,100 Web servers went down and the Gateway system was inundated with 1,100 messages. Fortunately, because such contingencies had been foreseen, the system was able to handle them.

Virtually No Training Required On The New System
When Logical Approach agreed to handle the telecommunications company's upgrade, its developers agreed to work closely with the employees during the entire process. The first part of the upgrade took about three months. Refining the system and adding features took another three months. The developers mimicked the older system as much as possible. Because of that, and because they had worked with the employees during the installation, the employees required virtually no training on the new system.

Was The Installation A Success?
Interestingly, the telecommunications company forgot to advise its people in the field when the upgraded system was brought into service. When a company conference brought all its engineering managers to its headquarters, virtually none of them knew the precise date of the cutover. Yet, each of the field engineers had deduced that the new system was online, only because complaints about the system had virtually ceased.