Generating Virtualization Application Profiles

Gaining the Empircal Evidence to Unveil Real End User Experience in a Virtualized Environment

Every mission-critical application that underpins a user’s primary duties is measured by Aternity’s Frontline Performance Intelligence Platform. Various users leveraging the same application in a different manner, automatically yield unique application usage profiles by the Aternity platform.

The Aternity Frontline Performance Intelligence Platform can rapidly generate and clearly visualize the key metrics, and their dependencies, of all the major components of a virtualized application environment, and thereby precisely describe the impact on end user productivity, as follows:

• CPU, physical, and virtual memory resources requirements linked to specific application processes. This is especially relevant for low-memory, low-CPU-consuming applications that could have high-server density, ensuring that each server is operating at an optimized capacity.  In addition, this is critical for determining whether peak CPU, IO, or memory resource consumption by an application will overburden a particular SBC environment
• Network latency, inside and outside the LAN, and application chattiness represented by the requests and responses passing through the client-side application and the back-end infrastructure.  Chatty applications that have a high-volume of latency-oriented communications with the back-end systems often benefit from closer proximity to the data center, as well as transmitting only screen scraped images
• Velocity of screen refreshes and the magnitude of rectangles updated by the application GUI is another key metric that should not be overlooked, especially when less optimized server-based computing protocols are involved. High velocity of screen refreshes would require large bitmap transfers, and in the case of an SBC virtualization implementation would not be recommended
• Acceptable latency thresholds determined by performing real-time analytics on the end user experience metrics, and correlating them with actual user productivity.  This will provide enterprises with the insight required to make fact-based decisions around resource allocation, including when to move or re-tune underperforming or underutilized virtual machines once deployed in production
• The number of application requests and responses, correlated with their impact on the latency of average process execution start-up, e.g. CPU bursts, I/O bursts, etc., will help determine the optimal virtualization strategy for a particular application mix
• The impact of local storage I/O (excluding page file) and network storage I/O are critical to understanding the impact of a virtualized environment upon end users
• The mapping of virtual machines to physical servers, their density, and the interplay of the different virtual machine images across application mix, time and configurations

Every measured parameter should include some relation to the peak and off-peak hours of activity, as the average alone is not sufficient. This is extremely important in order to make sure that the VDI can withstand the average peak hours’ load. One way to accomplish this is to determine when the peak hours are, and measure averages for both peak and off-peak hours.

Comprehensive analysis of the above metrics will result in empirical evidence of how users will actually

For certain scenarios the results might indicate that a “legacy desktop solution” would be more adept for handling the type of load imposed or the performance expectations of their user community. Examples of such include stock trading and Computer-aided Design (CAD) applications, and this should be taken into consideration from the outset.