An unstable site is one that begins to experience aberrant behavior or degraded performance as traffic increases. Often, these sites are thrown into a state that they do not recover from even as usage wanes. Some of the largest and most respected retail brands have encountered these issues at one time or another and often during the holiday peak season. We specialize in fixing these problems.
Having founded the company with an emphasis on this service, Tacit Knowledge maintains a suite of diagnostic tools and associated troubleshooting process that has worked in every engagement – no exceptions.
The process incorporates a logical, reductionist approach to problem isolation that is far more efficient than gut-based or ad hoc trial and error methods often used. It is also iterative, with at least one modification pushed to production on a daily basis. Our approach to unit testing, software quality and deployment automation are decisive in enabling this pace of change. With experience in the full lifecycle of commerce software development and deployment, Tacit Knowledge engineers leverage their combined development and operational expertise to analyze all aspects of the system; software, configurations, supporting technologies, and environments.stress-to-failure and endurance. Each of these tests further exposes unique types of traffic-related problems.
The end result is efficient software, verification of peak capacity and a scaling model used to drive any expansion strategy.
Application performance testing is often viewed solely as a pre-site launch activity to offer assurance the software will work in a live environment. In addition to this baseline objective, we test to predict the capacity of the system and to identify future scaling bottlenecks. It is also true that optimized software will reduce present and future costs by making the application more efficient.
This effort needn’t be extremely costly, and the approach is two-fold. First, we execute iterative performance tests throughout development. This mitigates end-of-project schedule uncertainty and identifies performance issues early. All tests use realistic amounts data including orders, categories, products, and users. These iterative tests apply smaller amounts of load to under-sized hardware to isolate several classes of performance and stability issues.
Second, a software release candidate is subjected to three different kinds of tests: peak load,