Virtualization is becoming increasingly ubiquitous in today’s datacenter. It enables multiple applications, including those running on different operating systems, to execute on a single machine, without any modification. Evolving from a mechanism to enable easy development and testing of applications, virtualization is now effectively used across the entire application lifecycle to reduce hardware costs, minimize space and power usage and to control support costs.
It is only natural that there is now a desire to apply these same savings to applications at the “edge” such as financial transactions and process control where the Stratus® ftServer is extensively deployed. In this scenario however, application constraints and the deployment environment impose a whole different set of requirements.
The key to effective virtualization is understanding where applications are under-utilizing the hardware. In many instances, applications use 20% or less of a machine’s capacity, and as processing power increases with successive generations of technology, this figure continues to drop. Some applications only need to run at specific times of the day, leaving processing capacity unused for the remainder of the time. Allowing for the overhead of the virtualization software itself, this enables a multiple applications to be hosted on a single machine.
However, the “edge” world is not so simple. For many ftServer applications, predictable performance is a critical requirement. Applications, such as financial trading, payment processing, some areas of process control and data acquisition, operate within tight parameters that make them unsuitable for virtualization. If performance critical applications compete for processing resource at the same time, the result can be disastrous. This is because applications running in a virtualized environment believe they have the whole processing capacity of the machine to themselves when they are executing. In typical IT applications, such as web servers, a short delay while processes wait to execute is generally not a problem. For critical edge applications, competition for resources, and the resulting delay, can be deadly.
A further challenge in edge scenarios is the increasing need for “always on” performance in remote locations, where there is minimal IT expertise. This further complicates the process of virtualizing the environment. It is important to note that while any single application failure may not be deemed critical, as more and more applications are virtualized onto a single machine, the impact of all of them failing at the same time can be catastrophic. This means that preventing downtime becomes an increasing priority in virtualized environments.
While it is true that virtualized environments can be configured to support availability, most solutions generally involve complex hardware and cabling configurations, often involving external storage arrays, and complex scripting to effect the failover. And with all of the complexity, you still end up with downtime in the order of minutes while the system recovers from a failure.
At Stratus we are seeing increasing interest for ftServer in availability sensitive environments such as the Edge. Why is this? Because it provides an always-on capability with simple operation. Since failover is instantaneous, and internal to the ftServer, failures are completely transparent to the applications and their users. Through Stratus support services, any failed components can be automatically detected and replaced without any knowledge or impact to the user. The simplicity of the solution, the use of a single piece of hardware and the minimization of the number of software component licenses required makes ftServer a very cost-effective solution.