Shunra

Any IT organization seeking to cost-efficiently deliver consistently high service levels across the enterprise must be able to accurately model the production network environment. Accurate network modeling is the key to effective testing, capacity planning, diagnostics and service-level assurance. If you can't accurately model the production environment, your ability to analyze current and predict future application behaviors will remain limited at best - especially when it comes to understanding the quality of the end-user's experience at all of your geographically dispersed locations.

Unfortunately, many IT organizations rely on tools that use mathematical calculations alone to model the production network environment. These tools require engineers to input a complex set of parameters into the software, which then uses those parameters to build a theoretical model of the environment.

This type of purely mathematical network modeling is prone to error for three reasons:

1) It assumes that the behavior of IT environments can be accurately predicted with parameters defined by the software vendor.

This assumption is not reliable. In the real world, networks and applications often behave in unexpected ways. Network devices may not be ideally configured and software code may have "bugs" of varying severity. Purely mathematical network modeling cannot fully take into account these real-world vagaries. Also, as the number of parameters required for the network model grows, so do the number of assumptions that must be made about the enterprise environment – and the likelihood that inaccuracies occur. IT organizations that rely entirely on mathematical network modeling therefore expose themselves to significant risk.

2) It makes the accuracy of the network model entirely contingent on the skill, precision and completeness with which an engineer determines the appropriate values for the parameters provided. This introduces multiple potential sources of error. Engineers may not always input realistic or accurate values. Conditions on the network may have changed in unexpected ways. The nature of new or newly modified applications may not be adequately understood. Any of these sources of error will cause a purely mathematical model to yield inaccurate analyses and predictions.

3) It requires constant updating to keep parameters in line with the ever-changing enterprise environment. Essentially, the IT organization winds up having to maintain two separate environments: the production environment and the modeled one. Keeping these two environments in sync is a major challenge. If they're not kept rigorously in sync, the model will fail to yield accurate and reliable results.

Shunra's solution to this problem is to take an empirical - rather than a purely mathematical - approach to network modeling. Shunra's empirical approach provides IT organizations with a "Virtual Enterprise" that reflects conditions in current and projected production network environments with unmatched accuracy. It is thus far more effective for discovering application and network performance problems, determining infrastructure readiness, and assuring service-level compliance to the end-user's desktop.

Shunra's empirical approach is substantially differentiated from purely mathematical approaches by two key characteristics:

Empirical Input from the Production Environment

Under Shunra's empirical approach, the behavior of the product network environment can be captured and imported from the production environment, rather than constructed using purely mathematical inputs. Shunra records up to 30 days of network conditions on a 24x7 basis, and uses that data to define parameters such as latency, error rates and bandwidth utilization. This approach eliminates guesswork and ensures the accuracy of the network model. It also allows "what if...?" scenarios to be created by making incremental adjustments to an accurate baseline network model - rather than by speculating about possible future changes in environmental behavioral parameters.

The Use of Live Applications and Production Data Center or Lab Resources

Instead of relying exclusively on abstract mathematical formulas to model application and network behaviors, Shunra's empirical modeling environment uses the production data center servers or lab servers, and live applications to drive scenarios. By doing so, Shunra'sPerformanceSuite reflects the specific idiosyncrasies of real-life applications and servers much more precisely than any set of fixed "point-in-time" application parameters ever could. This ensures that scenarios accurately portray the quality of experience at the end-user's desktop.

In addition, because this empirical approach does not depend on the input of sophisticated mathematical parameters by network engineers, it enables developers and QA staff to take advantage of modeling throughout the design and coding process. This is a powerful benefit, since it permits IT teams to discover early in the lifecycle any potential problems relating to the ultimate performance of applications and infrastructure in the real world. The applications can therefore be fixed prior to production roll-out, saving significant staff time and money while ensuring that delivery deadlines are consistently met.

This doesn’t mean that mathematical network modeling has no place in IT. There are situations in which real-world applications, data center resources and/or network infrastructure are not available for use in an empirical model – or where the scenario being modeled simply diverges too greatly from existing norms for empirical network models to be fully effective. In these specific cases, mathematical models are obviously applicable. But mathematical models can never deliver the accuracy and reliability of empirical ones.

Thus, IT organizations that want to ensure reliable, consistent quality-of-service for critical applications across the distributed enterprise should embrace empirical network modeling and apply it early and often in the application lifecycle. By applying empirical network modeling in the form of Shunra's PerformanceSuite, IT can keep development projects on track, more effectively plan for future infrastructure requirements, accelerate troubleshooting, optimize the end-user experience, and eliminate the risks of introducing inadequately tested applications into the production network environment.