Enterprise Network Application Services Optimization
US Army Network Enterprise Technology Command (NETCOM) is continuously exploring how to deliver IT services in the most effective manner to their community of users. As a globally distributed organization, they evaluate the technical and operational tradeoffs between different deployment architectures. A series of investigative projects were authorized that were estimated to cost approximately $800k and take six months to accomplish if a physical test environment was utilized. SCALABLE executed on a number of tasks ordered awarded by NETCOM that focused on the evaluation of different technical architectures for secure communications.
They enlisted SCALABLE to help them determine whether centralized or decentralized application services would be the most effective in addressing their needs. They were interested in finding out answers to the following questions:
- What is the optimal architecture for delivering Online Certificate Status Protocol (OCSP)/Public Key Encryption (PKI) services?
- Would migration from a VPN environment to an MPLS implementation offer better user performance?
- How could Microsoft SharePoint Services be tuned for increased user experience?
NETCOM chose to work with EXata and was able to create a high-fidelity virtual network model which was then integrated with specific live servers, switches, routers, and end-stations via a "system-in-the-loop" interface. A wide range of different network geometries, failure modes and application flows were examined. Engineers were able to identify a network configuration that optimized the OCSP/PKI performance and cost. Other scenarios answered the question of whether VPN architecture provided better performance than a network-wide implementation of MLPLS.
Further use of the virtual network models, combined with actual packet capture of SharePoint traffic patterns, enabled NETCOM to evaluate multiple options for tuning the infrastructure and software configurations and to identify "optimal" network constructs for application support. The overall simulation exercise took less than two months and cost less than $300k.