Robust or Go Bust

Leading-edge E-businesses are finding that they are not only blazing new trails with their business models, but also with their IT infrastructures. At first glance it may not seem obvious that the young Internet start-up is going to face challenges that are different and in some ways more extreme than their established brick-and-mortar brethren, but indeed this is the case.

While most established firms have had the relative luxury of building their now solid IT infrastructures over a protracted length of time, dot-coms are finding themselves needing to build these infrastructures overnight. The normal metrics for speed, scalability and capacity planning no longer seem to apply. Instead, dot-coms are finding themselves needing to build scalable applications that often go beyond those of established firms, but needing to do so in a fraction of the time, while dealing with unprecedented volumes of data that often descend on their fledgling enterprises in unpredictable ways.

Consider the following examples:

Princeton eCom, a leading provider of electronic bill payment and presentment (EBPP) solutions over the Internet serves as an outsourced EBPP solutions provider to companies like Bell Atlantic Mobile, Adelphia Cable, and Cinergy Corp. Princeton eCom is a leader in the fast-growing on-line bill presentment and payment space. The service-level agreements with their corporate customers are demanding – all bills are typically presented to the customer within four hours of receipt of the date, regardless of the time files are received. Princeton eCom needs to be continuously ready to process the incoming data since there is no way of knowing when the data will arrive or when customers will want to access their accounts. A failure to meet these stringent agreements results in financial penalties.

The National Association of Securities Dealers (NASD), parent company of NASD Regulation and the NASDAQ and AMEX stock markets, recently implemented an Internet-based regulatory application (Web CRD(sm)) providing broker/dealers and regulators with on-line access to reports from a database containing records on more than 5,500 broker-dealers and 600,000 brokers. The reports can be accessed electronically through the end user’s browser, or downloaded. The system currently generates and distributes an average of 500 reports per day, many of them running to thousands of pages in length. Yet, while IT can predict general usage patterns (e.g., more reports will be requested during trading hours than late at night), the system must be prepared to handle requests far in excess of its average capacity, any time of the day or night.

Tosco Corporation, one of the largest independent petroleum refiners and retailers in the United States, has been in the process of Webifying its ERP system. Besides making reports available to remote users via the Internet, the company is leveraging ERP information to present select customers with browsable, secure extranet Web pages. Buyers have on-line access to information such as pricing which can be incorporated into their own information systems. While customers can be presented with a set of pages that are customized to their specific needs using site content management software, process management is still needed to make sure the right data gets posted to the right customer-specific page at the right time.


These examples point to some of the unique challenges that must be addressed in order to ensure that an E-business environment remains available, reasonably responsive, and reliable – and thus, successful in the long term:

Data Volumes: Dot-coms and click-and-mortars alike are finding themselves dealing with vast amounts of data. It wasn’t that long ago that the world’s electronic data comprised less than a few terabytes. Now individual corporations store terabytes and receive gigabytes of new information daily. One Internet advertiser is known to add more than seven million rows to its Oracle database daily.

Unpredictability: Further complicating the problem, the immense volumes of data do not arrive at predictable times or in predictable quantities; to the contrary, they may arrive all at once, packing the vicious punch of a “data tsunami”. Such tsunamis can be random events; at other times, companies unwittingly bring them upon themselves (for example, by undertaking a major Web-oriented marketing initiative, and failing to clue in IT). Such sudden influxes of data can easily overwhelm even robust, high-performance, high-capacity environments. They also complicate the task of scheduling and automating processes, since companies don’t know when the data will be arriving and when the batch processing will need to begin.

Nonstop Processing: The unpredictability of the data flows have led to the need for just-in-time processing. While the data centres of yesterday were able to do traditional batch scheduling, the unpredictability of incoming data and today’s insatiable need for speed and customer responsiveness have created the “real-time” organization where all processing needs to be done as soon as possible. There are no off hours and the enterprise needs to be open and ready for business all day, everyday, without fail, no matter what the processing load.

Continuous Availability: In the Internet environment, customers have come to expect total, 24-by-7-by-365 availability with reasonable response time (from the end-user perspective, a system that is available but unable to respond to his requests in a timely manner is little better than one that has crashed). And when their expectations are not met, end users need only push a button to be connected with a competitor. “We have CFOs check our site at 2:00 a.m.,” explains a manager at Princeton eCom, noting that these CFOs are merely seeking to ensure that their own customers are getting the service they promised, and Princeton eCom contracted to deliver.

High Visibility: Adding to the perils of e-commerce is the visibility of a company’s internal operations to the world at large (and customers in particular). In a traditional IT environment, system outages typically went by unnoticed by all but the few directly affected (usually internal employees). Today, when high-profile companies like eBay or Amazon go down for an hour, it makes international news. And, while the failure of a more specialized system like the one at NASD might not make the headlines, it would certainly be noted by a broad array of the company’s customers, resulting in disgruntled end users and, potentially, lost revenues.


Businesses today need to plan aggressively if they are to build well-performing systems capable of handling the stress and strain of today’s fast-paced and unpredictable world. We have identified four key areas that deserve attention when looking at building a robust Internet infrastructure.

Plan: Although the inherent unpredictability of E-business makes advance planning challenging, it is possible to broadly define requirements. Explains Laura Davidson, assistant director, office of technology services for NASD, “You have to know what your workflow is going to be like and what the volume of activity is going to be like and how the systems are going to be stressed so that you don’t set unrealistic goals. Start with small steps, and then make it better after you get it to work.”

Overbuild: The most common solution to challenges of E-business has been to simply throw more resources – human and hardware – at the problem. While such an approach is not sufficient to meet the problem, it is probably necessary. “We take every single point of failure very seriously,” says Princeton eCom CIO Don Brenton. “The investment in infrastructure is probably more costly, but then again the ROI we work on is availability ROI.”

Automate: Unfortunately, no amount of planning and infrastructure investment can guarantee continuous processing when the data tsunami hits. Back-end automation is absolutely essential to achieve the efficiency required to handle the escalating data processing requirements, and to ensure continuous processing in accordance with business rules when the system is stressed beyond capacity. An E-business like Princeton eCom would be impossible to administer without such automation tools, particularly given its zero-tolerance for downtime, with customers expecting almost immediate balance updates once they’ve paid their bills. States Brenton: “In our quest to eliminate manual processes we have searched for and implemented a variety of tools. We will continue to pursue automation vigorously to improve our performance. These tools also lessen our time to recover from would-be operational problems. They also lower our operating cost.”

Monitor: Like many E-businesses, Princeton eCom employs an outside agency to monitor its Web environment and deliver independent reports on performance and availability, complementing its in-house monitoring tools. The company has defined thresholds, and established action plans tied to those thresholds to ensure that capacity and performance can be increased ahead of user demand.


The dynamics of today’s E-business environment have created the need to rethink traditional process management. Previously the business would conduct much of its recurring business processing at night during the so-called “batch window” when the majority of the systems were off-line. In today’s world, however, the batch window is all but gone and there is a compelling need for nonstop business process management that is capable of responding in near real time to changing business conditions.

Until recently, even in complex, heterogeneous environments, scheduling has come down to a more or less “static” ordering of predictable data-processing events, (i.e., generating invoices every Tuesday at 2:00 a.m.), typically for an internal user community. The E-business environment is far more dynamic, with a steady stream of events. Yet total on-line processing is simply not feasible for the vast majority of E-businesses. “There’s no way you could build an environment large enough to handle every transaction as it comes over the network,” says NASD’s Davidson. “We knew we had to manage the workloads.”

The solution is to incorporate process automation software that is capable of handling both static scheduling and dynamic scheduling. NASD takes this approach for managing its Web-based reporting system. Incoming requests are gathered and, depending on the availability of resources and a variety of other policies, the requests are scheduled and processed. The system serves as a control point for managing the processing of the requests and the balancing of the workload to ensure reasonable response times for other applications.

Dynamic process automation also has the potential to increase operational efficiency. This requires using automation tools that can interact with the data and output from the application, which would include being able to read application data directly from an application’s underlying RDBMS. For example, a major hotel chain with an on-line reservation system has found that running a particular batch job on fewer than 100 reservation requests results in inefficient processing – yet it can have no way of knowing when customers will make hotel reservations. The solution involves establishing a “trigger” to automatically run the job when a sufficient number of reservations have come in (subject to availability of processing resources and other dependencies). Another example of dynamic automation is having tools monitor the output data files or reports and check for conditions. For instance, if the debits and credits don’t balance on the ledger reports a tool should be able to ascertain that directly and take corrective action, as oopposed to having to wait for a manual review of the data and the attendant delay.

Faced with exponentially increasing volumes of data and the need to ensure continuous availability in the face of unpredictable demands, E-businesses must plan ahead to ensure robust, efficient workflow. A tool for dynamic workflow automation can assist in this process, maximizing the efficiency of computing resources and serving as a single point of integration between an e-commerce application, back-end ERP and financial applications, and report delivery tools. Simply stated, a failure to automate will put the business at risk, expose holes to the outside world and drive customers to competitors.

William (Bill) Wrenn has served as President and Chairman of AppWorx Corporation since its inception in September 1990. He was IT Director for Airwork Corporation for three years, and previously spent over 13 years serving in IT management with General Electric Company