Slowmotion Benchmarking: Measuring Thin-client Performance

Thin-client computing offers a solution to the increasingly unmanageable costs of today’s IT infrastructure. Thin-client computing moves user data and application management from the desktop to centralized servers in professional-managed data centers. Application service providers (ASPs) are using thin-client computing to enable desktop machines, web browsing terminals, and other low-cost embedded devices to function as simple user interface devices for accessing computational services over the Internet. ASPs have the potential to deliver easier-to-maintain computational services with reduced total cost of ownership.

A key enabling technology in thin-client computing is the remote display protocol. The protocol allows graphical displays to be served across a network to a client device, while applications and even window systems are executed on the server. Using such a protocol, the client transmits user input to the server, and the server returns screen updates to the client.

To assess the viability of the thin-client computing model, we are conducting a series of detailed experiments to quantify the performance of thin-client platforms for various application workloads. We focus on the network performance of remote display technologies. Our experiments include measurements of popular thin-client platforms such as AT&T VNC, Citrix Metaframe, Microsoft Terminal Services, Sun Ray, and Tarantella. To measure thin-client performance even in the case of proprietary, closed systems, we have developed a novel, non-intrusive measurement methodology called slow-motion benchmarking, which has resulted in the NCL ThinBench benchmark suite. Our results indicate that current thin-client solutions generally work well in LAN environments, but their performance degrades substantially in the broadband environments envisioned by ASPs.

To address this problem, we are developing new remote display technologies that will enable ASPs to deliver graphical and multimedia applications in a broadband environment. We are also investigating issues in scalable server resource management to support the demands of large numbers of interactive users in thin-client environments.

More Information:

Other Resources:

Columbia University Department of Computer Science