GFT AG

GFT AG designs and implements IT solutions with a focus on the financial services sector. By combining extensive project and technology experience with in-depth industry expertise, GFT serves as a strategic IT partner for renowned companies worldwide. GFT places IT specialists with companies across all industries. In the area of third-party management, GFT handles the complete management of IT service providers. GFT inboxx sets standards with innovative products in the areas of business process optimization, document management, and digital archiving. The GFT Group has a presence at 20 locations in seven countries. Its IT solutions are deployed in more than 30 countries. GFT AG currently has approximately 1,000 employees and 1,300 freelance IT specialists.

Server Sizing for Server-Based Computing

As part of a technology refresh, we carried out a project on behalf of GFT AG with the goal of identifying the most powerful hardware and software platform for a server-based computing platform (Citrix XenApp) for one of Germany’s largest banking institutions. The bank provides tens of thousands of users with access to all applications via a shared Citrix desktop using several hundred Citrix servers. The bank had previously selected Intelligent Solutions as its partner to carry out the project.

User Experience as a Performance Metric

The performance metric was the number of users per server while adhering to existing service SLAs regarding application response times, which were measured based on end-user perception at the user workstation.

The first project began with an analysis of the most frequently used applications, weighted by the resource load they generated. From this, a representative application portfolio was distilled, and for each application included in it, the most frequent/important business transactions/workflows were identified in consultation with business users.

These workflows were mapped for automated execution using the CitraTest VU^® test automation solution. The resulting execution sequences were then modeled—in terms of composition, number of instances, and intensity—into an execution scenario that, as the number of users increased across all subsystems, exhibited the same resource consumption behavior as production operations (scenario calibration). The determined scenario was used to run stress tests against the hardware and software configurations under evaluation. Application response times visible on the end-user’s screen were also measured using a software robot (CitraTest^® APM).

The following questions had to be answered as part of the hardware refresh of all Citrix XenApp servers:

• Which hardware configuration, when combined with which operating system configuration, delivers the best performance? Different numbers of CPUs, CPU and front-side bus clock speeds, various I/O controllers and hard drive models, different amounts of RAM, paging file sizes, RAID configurations, the effects of security and monitoring policies, and local vs. roaming user profiles resulted in over 70 possible server configurations!
• How well do different virtualization scenarios perform for Citrix XenApp servers?

Reducing the number of test runs through differential analysis

There was not enough time remaining before the decision deadline to test all possible hardware combinations. Therefore, the configurations to be tested had to be selected in such a way that, with the minimum number of tests, it would be possible to reliably infer the performance of the untested configurations. Overall, for the first research question, we were able to reduce the number of configurations to be tested to 20 representative ones and determine the effects of various influencing factors on bottleneck situations using the SightLine performance analysis solution.

The project had a turnaround time of 50 person-days from commissioning to the delivery of the requested findings. Based on the project results, significant, financially impactful efficiency gains were realized.

Virtualization Strategies for the Server-Based Computing Environment

Although hardware and software releases had changed significantly in the two years following the initial project, the calibrated scenario continued to accurately replicate the actual load behavior in production. The second research question could therefore be answered with fewer and shorter project steps within an 8-day project duration:

• Adapting and testing the measurement procedures to account for user interfaces that had changed over the course of previous releases
• Configuring and executing multiple measurement runs against different virtualization scenarios
• Comparing scenario performance and identifying the causes of the observed differences
• Preparing a report with test results and recommendations for the optimal configuration

The project results formed the basis for decisions regarding the design of the virtualization strategy for the bank’s server-based computing environment.