Feature

Case Study: Intel’s Fast Ethernet solution

Intel’s Fast Ethernet solution helped US credit card processing company First Data Corporation improve performance and preserve uptime

With more than 700 clients and yearly revenues of $3 billion, First Data Corp. (FDC) handles 60 per cent of all the credit card transactions processed in the United States. The large amount of transactions required per second makes imposing demands on information technology.

In the past, those millions of transactions have been processed by a large mainframe computer at corporate headquarters in Omaha, Nebraska. Today, FDC is involved in a five-year re-engineering project to move credit card transaction processing to distributed Intel-architecture systems running Windows NT. The Card Services Group, a 90-person development group located in Marlton, New Jersey, is responsible for creating the new 24-hour customer service application that will shift the company to inexpensive and easy-to-use PC systems.

Paul Nelson is a senior systems administrator for the company's Card Services Group. His key role in the re-engineering effort is making sure in-house application developers have access to the resources they need. A year ago, that challenge seemed insurmountable." When I arrived here in March 1995, I inherited a single-collision 10Mbit/s Ethernet domain," Nelson recalled. "We had NICs that would routinely start broadcasting and then shut down the network because the card would go bonkers."

Since then, the Card Services Group has been migrating to a combination switched and shared 100Mbit/s network by deploying a Fast Ethernet workgroup solution based on products from Intel Corp. The solution includes Intel EtherExpress(TM) PRO/100 LAN adaptors installed in servers and clients, an Intel Express Switching Hub and an Intel Express Stackable Hub.

The Fast Ethernet advantage

"Since putting the EtherExpress cards in, we've provided a much more reliable service to the Card Services Group," Nelson reported. "One of my goals was 99 per cent uptime during business hours. I've achieved it with the help of the Intel cards. And my users are no longer complaining about server lag."

With fewer interruptions from users experiencing server problems, Nelson can spend more time developing his network and less time troubleshooting.

Performance has also improved. Network utilisation, for example, has dropped from 60 per cent to 10 per cent. A significant improvement has also been achieved when performing backups of local user workstations. Using internal tools in Windows NT, Nelson determined that maximum byte throughput during workstation backups jumped from 496,000 to 695,000 bytes per second ( an increase of nearly 50 per cent.

Three factors that influenced FDC to install Intel adaptors rather than other vendors' offerings were ease of installation, scalability and attractive pricing. Nelson has this to say about installation: "I plugged the card in, plugged the driver in, it detected the network and we were off and running on a 10Mbit/s-switched backplane."

Because the EtherExpress PRO/100 Adaptor supports both 10Mbit/s and 100Mbit/ps, desktop systems could initially operate at the slower speed. Then, when Nelson swapped in the Express Switching Hub ( a 10-minute installation ( it was a simple matter of turning each system off, then turning it back on. The adaptors automatically detected the new hub speed and have operated flawlessly at 100Mbit/s ever since.

"The price was right, too," Nelson said, noting that an EtherExpress PRO/100 Adaptor is comparable in price to a 10Mbit/s card. "For the extra $30, I potentially bumped my bandwidth 10 times." As for the Express Switching Hub, Nelson added: "It not only has been working fine, but it has been in service since we installed it. We've tested it extensively and had no problems with protocols."

Nelson most recently installed his first Express Stackable Hub to round out Intel's Fast Ethernet workgroup solution. The initial installation involved plugging it into the switch and then plugging users into the hub-taking all of five minutes.

At present, all of the servers and the majority of the desktops in the Card Services Group are running at 100Mbit/s. Eventually, all clients will be on Fast Ethernet, but the migration will continue to occur in stages to preserve FDC's investment in legacy equipment.

With the ease of installation built into the Intel products, moving a workgroup to bigger bandwidth has become painless. Nelson simply takes down a section of the network during lunchtime ( and tells users they'll be running at 100Mbit/s when they get back.

"It's an incremental and inexpensive way to easily migrate from a 10Mbit/s network," he said. "I don't have to tell management we need to spend $50,000 to upgrade all users now."

Implementing the solution at client sites

Armed with a solution that had proven itself in-house, FDC, in late 1995, began deploying the new customer service application at target sites that are clients of its customer service division. Two client sites-banks in Mexico and Tulsa, Oklahoma, are already live and running, with users giving high marks to the graphical interface that replaces the arcane text-based sessions. Eventually FDC hopes that the customer service application can be implemented at all 700 of its client banks throughout North America.

Nelson said that every new system deployed in that plan would have the Intel EtherExpress PRO/100 Adaptor installed to help customers move easily and affordably to Fast Ethernet. "Target banks can't afford to spend the $500 per system it would take to upgrade their existing Ethernet or Token Ring cards to use FDDI, and there are too many issues related to ATM that haven't been resolved yet, like the fact that standards are not in place," he explained. "On the other hand, it's very easy to migrate them to a 100Mbit/s Fast Ethernet solution. Fast Ethernet allows you to upgrade bandwidth with proven technology in a scalable fashion."

Rollback plan gathers dust

Throughout the network migration to Fast Ethernet, Nelson had a rollback plan in place. He formulated the plan so that if, for any reason, a new network addition didn't work immediately, he could confidently reinstate the previous network configuration. To date, none of the Intel products have called the plan into action.

"Intel cards, hubs and switches have proven reliable on systems that need to remain running on a 24-hour basis," Nelson concludes. "Network uptime is a key issue for us. My bonus depends on it."

Compiled by Mike Burkitt

(c) 1999 Intel Corporation


Email Alerts

Register now to receive ComputerWeekly.com IT-related news, guides and more, delivered to your inbox.
By submitting you agree to receive email from TechTarget and its partners. If you reside outside of the United States, you consent to having your personal data transferred to and processed in the United States. Privacy

This was first published in January 1999

 

COMMENTS powered by Disqus  //  Commenting policy