The case studies presented here are real stories from our experience. Unlike fairy tales, they don't all have happy endings.
Anticipatory Request Strategy
A new system with a gorgeous, highly functional interface based on rules technology was intended to replace a system where the rules were “in the users heads”. The system proved to be unacceptably slow after the install.
The architecture did not support speeding up the middleware pipes. We analyzed the pattern of requests to the servers and implemented a strategy of making requests to the servers in anticipation of the user clicking buttons. By the time the request was made, the system already had the data.
Despite the additional processing that was required for the new interface, we were able to set up side-by-side comparisons with the legacy system and prove that the new system was not only more functional but faster. The speedup was attained in 3 months.
Everything Right
Weekly time entry system for temporary staff proved unstable under heavy load. Almost every Monday morning, when people had to enter their time so they could get paid for the week, it would crash or hang. The situation was stressful for all concerned and was a threat to the business.
Management decided to build a new system because it was perceived as unfixable. Despite the short timeframes — 8 months from funding to install — the performance requirements were defined early and monitored often. Other business goals (good-looking & functional interface, all the required features and yes, even scope creep) were also respected.
The system has far exceeded its design goals and the sales team feels no constraints when signing up new customers.
Headed for trouble
A high visibility web services project appeared headed for trouble when Dimensional Modeling showed that the system would not scale.
Given the criticality of the project, it was decided to install as scheduled and address problems afterwards. The product definition had to be scaled back after the install.
The business impact — in dollars as well as the opportunity cost of missing the product window — was significant. The new changes were partially installed when a new management team decided to cancel the project.
Sigh of relief
There was a concern that the system would be OK when installed but not be able to keep up with the anticipated growth in the business.
The workload was carefully modeled and extensive measurements were taken. In the process of making the measurements, two configuration issues were found, fixed within days. A third system bottleneck was identified as something that would start to manifest after about a year of growth.
As long as the change could be made within a year, the system would continue to scale into the foreseeable future. The change was scheduled into the product development calendar.