[A Cisco Transaction Connection and StarSQL Enterprise Edition Case Study] Employers Reinsurance Insures Fast, Global Access to DB2 Data Employers Reinsurance Corporation’s implementation of the Cisco Transaction Connection and StarSQL data access middleware has resulted in significant cost savings, dramatic performance improvements, greater scalability, as well as reduced system overhead and maintenance. Employers Reinsurance Corporation (ERC), whose parent company is GE Capital Services, is one of the largest and fastest growing reinsurance organizations in the world. ERC needed to eliminate redundancy, ensure consistent results across the organization, and provide its vast user community with fast, efficient access to nearly a terabyte of crucial IBM DB2 data. Identifying the Problem Headquartered in Overland Park, Kansas, ERC has offered global risk management for more than 100 years in over 40 countries in key support services, including legal, actuarial investments, financial services, and underwriting. ERC’s rapid growth and recent string of acquisitions resulted in the need for a standard method of handling transactions. "After the acquisitions, we had multiple systems doing the same tasks, so we had a lot of redundancy," recalls ERC Vice President and Chief Information Officer Henry Slack. "We wanted to develop a common system across the company. We knew if we could accomplish this that we would reach a much higher level of productivity." To meet the challenge, ERC used a two-tier client/server model. The server system uses IBM’s DB2 relational database on an MVS (Multiple Virtual Storage) operating system on an IBM mainframe. The clients are PC desktop systems running Windows NT Workstation operating system. The network is comprised primarily of Cisco routers. Enterprise-Wide Access "What we wanted was a central data server, because we didn’t want to distribute data all over the world," Slack explains. "It’s much too difficult to replicate this amount of data and try to keep 15 different locations in sync. We felt that the low-cost alternative was to have a large database server-in our case MVS and DB2-centrally located at our headquarters, and then distribute our clients out." Slack states, "We wanted as few moving parts as possible-the more intervening places you have to stop, the more you introduce delay and increase the chance for failure." A key component of ERC’s solution is database access middleware needed to gain access to DB2 data. ERC looked at various solutions to connect its NT desktops, located worldwide, to its centralized DB2 server. Most of the solutions that ERC evaluated were expensive, lacked scalability, were hard to install and maintain, and difficult to manage. They found that UNIX gateways required regular load balancing and constant monitoring. Handling Gateway Bottlenecks "We started development with a product that used UNIX systems as gateways to DB2," says Slack. But when it came time for Slack’s team to deploy the solution into production, it was clear that the UNIX gateways were in fact bottlenecks and could not scale to the load. "The cost of additional gateway systems, combined with lackluster performance, led us to consider alternatives," Slack recalls. After considerable due diligence and testing of the leading connectivity products on the market, ERC opted for a powerful new enterprise-level connectivity solution from StarQuest and Cisco Systems-StarSQL Enterprise Edition (EE) and the Cisco Transaction Connection. Putting DRDA Standards to Work Jointly developed by Cisco and StarQuest Software, the Cisco Transaction Connection uses IBM’s Distributed Relational Database Architecture (DRDA). The Cisco Transaction Connection switches DRDA over TCP/IP LAN environments to DRDA over IBM SNA hosts. The requirement for matching network protocols on the LAN and host is decoupled and the requirement for OS, DB2 and TCP/IP upgrades is eliminated. TCP/IP on the Desktop Says Slack. "StarSQL EE does all of its processing on the client, which eliminates bottlenecks in processing at the Gateway or Mainframe. When we started using the StarQuest product, we noticed an immediate improvement in speed," Slack recalls. "When we added the Cisco Transaction Connection, we connected our TCP/IP desktops to our SNA backbone and the performance improved even more. If someone had claimed to provide a 400 percent performance improvement, it would have stretched my believability levels, but that’s what we saw for some of our applications." Eliminating FEP and Saving Millions The end result of ERC’s development efforts included closing two data centers, eliminating host-based 3270 applications, resolving year 2000 issues, plus improving performance and bringing consistent results to its entire user community. In Conclusion According to Slack, ERC has saved millions of dollars on this consolidation and rejuvenation of their applications. "Just the ongoing operational costs of keeping our disparate systems operational exceeded the purchase cost of our new system," he comments. "With this new server-based deployment," Slack adds, "we are in essence getting rid of seven major systems and eliminating redundancy, while dramatically increasing our functionality."