Finally, client-server systems can deliver on those promises.

Client-server technology promises banks tremendous price-performance advantages over traditional mainframebased information systems.

That's nothing new. Client-server technology has been promising tremendous advantages for years.

What's new is that client-server technology has now advanced to the point where it can deliver on its promises.

The client-server technology required for many banking applications, especially management information and decision support systems, is now in the marketplace. This technology can and will support the very large data bases required for enterprise-wide applications at today's multistate, consolidated institutions.

Clients of ours have successfully implemented decision-support systems that extract and store tens of millions of records on client-server platforms.

One system now being implemented will process over 20 million records a month and maintain several years of history; eventually it will contain over one terabyte of data. Similar successes are being reported in other areas, such as branch automation and marketing.

Effectiveness in Question

The real issue today is not whether client-server technology is competent to perform banking applications.

It is how effectively this architecture meets specific business objectives in specific institutions. In other words, the decision to implement a client-server system in a bank today is a business decision, not a technology issue.

Today, as in the past, business decisions about technology require trade-offs between price and speed or function.

This principle is particularly true of client-server systems, which can vary tremendously in cost and capacity.

The term "client-server" describes an enormous range of configurations, from simple local-area network setups with a single PC server to enterprisewide solutions anchored by parallel-processing "superservers" incorporating 20 or more microprocessors and costing hundreds of thousands of dollars.

Network Speed

Like other systems' architectures, client-server systems can work only as fast as their slowest components. The limiting performance factor in most banking client-server systems is the speed and capacity of the network.

Network communication installations can vary from simple modems using standard telephone lines, to local-area neworks connected by direct cables, to wide-area networks using dedicated fiber-optic digital transmission lines.

Wireless remote technologies, some using satellites, may soon add an extra dimension, but their cost-effectiveness cannot yet be estimated.

To be cost-effective, banking applications systems should take the advantages and consultants of the new architecture into account. We cannot always port existing applications directly to the new architecture and expect to register significant performance gains.

Maximizing Data Traffic

Some client-server systems, for example, work in ways that maximize data traffic on a network. A client microprocessor requests data from a server, which in turn downloads a file to the client. The client device then processes the file and returns the information to the server, which updates the source file.

This technique works fine for simple transaction processing systems, where the data volumes passed between client and server are small. But for management information and decision support systems, where requests can include millions of records, the data traffic can produce something like information gridlock on even the most advanced networks.

A good decision-support system, processing large volumes of data, can embody techniques for retaining much of the processing load on the server. Instead of questing the server to download all the data to be analyzed, the client processor can ask the server to perform an analysis prior to downloading. This technique may reduce network traffic by several orders of magnitude.

Consider an example from a bank profitability management system. The task assigned is a stratification analysis to identify deposit accounts with large balances. The data base contains one million records in this category.

Passing these records intact to a client processor might generate as many as 40 million bytes to data traffic. By having the server perform the analysis and pass only the selected records to the client, a system can reduce the resulting transfer load to a few thousand bytes.

Cutting Equipment Costs

By reducting data traffic, these techniques reduce the cost of network and communications equipment required to achieve performance targets. They also keep the server processor busy and utilize the cheapest mixture of processing power appropriate for the particular task.

Other tasks may suggest a different strategy for optimal use. Calculation-intensive processing, such as those required for financial modeling with repeated "what-if" scenarios, may be performed best on client-based microprocessors, either standalone or clustered, connected to the server through a high-speed local-area network.

This technique allows analysis maximum freedom without draining the resources of the server. In this case, the constraint is not the network, but access to low-cost processors to perform arithmetic calculations.

Optimal Configuration Varies

The important point here is that the optimal configuration for a client-server system will vary with each institution and its applications needs. So will the expense. Both the system and its costs will depend on the amount of data that must be processed, the number of users, their locations, and their performance requirements.

If all users are located very close to one another, performing similar tasks, and using a relatively small data base, the systems design is simple and costs are low. With large data bases serving a widely dispersed user community, communication and performance issues become more critical. Every institution has its own unique cost-benefit ratio.

What does this all mean to financial institutions? It means that bankers can now take advantage of the new client-server architecture to implement important functions such as decision support. But they will have to do their homework thoroughly.

Implications for the User

It also means that a decision to implement client-server systems has numerous implications for the user's relationship with vendors, as well as with other organizations within the bank. In terms of selecting systems and vendors, there are some important things to remember.

First, look beyond simple price-performance ratios. These figures can be misleading. The published performance benchmarks on which these ratios are based may not apply to the processing that will be done on the system. This caveat is particularly applicable to the full-table scans frequently required by decision support processes. There are not yet any really useful benchmarks for this type of processing.

Second, focus on integration skills. Integration is what the client-server concept is all about, and integration skills are critical to success with client-server systems. It does a bank no good to have the fastest and most expensive hardware if the applications software cannot take proper advantage of it.

The most powerful processors cannot drive an overloaded data network faster than its inadequate capacity allows. And no system will ever serve a bank properly if it is too difficult for the intenede users to learn and apply.

Plan for Expansion

Finally, look for a migration path for expansion and upgrading. Good information has a notoriously addictive effect. The more you have, the more you want. And the more useful the information proves, the greater the incentive to distribute it across an organization, to put it into the hands of all the people who need and use it.

Systems will grow. If they don't, they must eventually be scrapped and replaced by systems that can. Whatever the direction a bank takes, it will need an upward path.

Within the bank's own organization, the demands on internal resources will change. Clientserver implementations place a premium on network and communications skills. They also require significant data base administration and support. User departments will have to have someone qualified and fully committed to the administrative role.

Users must also realize that they have entered an operational environment. A client-server system isn't something that can be turned off at night. Users will need to learn to work closely with systems people to ensure that the system is properly maintained, backed up, and protected.

Once a client-server system is properly implemented and supported, bankers will invariably find that they depend on it for a growing number of critical tasks. growing number of critical tasks.

And that's a good sign. It shows that we are at last turning the long-deferred promise of client-server architecture into reality.

For reprint and licensing requests for this article, click here.
MORE FROM AMERICAN BANKER