Does "thin client" mean "energy efficient"?

The thick client –a personal computer with integral disk storage and local processing capability, which also has access to data and other resources via a network connection – is accepted as the model for providing computing resource in most office environments. The Further and Higher Education sector is no exception to that, and therefore most academic and administrative offices are equipped with desktop computers of this form to support users in their day to day tasks. This system structure has a number of advantages: there is a reduced reliance on network resources; users access a system appropriate to their needs, and may customise “their” system to meet their own personal requirements and working patterns. However it also has disadvantages: some are outside the scope of this project, but of most relevance to the green IT agenda is the fact that relatively complex and expensive (in first cost and in running cost) desktop systems and servers are underutilised – especially in respect of processing power. While some savings are achieved through use of “sleep” modes and similar power reducing mechanisms, in most configurations only a small portion of the overall total available processor resource is utilised. This realisation has led to the promotion of an alternative paradigm, the thin client. In a thin client system, the desktop is shorn of most of its local processing and data storage capability, and essentially acts as a terminal to the server, which now takes on responsibility for data storage and processing. The energy benefit is derived through resource sharing: the processor of the server does the work, and because that processor is shared by all users, a number of users are supported by a single system. Therefore – according to proponents of thin client – the total energy required to support a user group is reduced, since a shared physical resource is used more efficiently. These claims are widely reported: indeed there are a number of estimation tools which show these savings can be achieved; however there appears to be little or no actual measured data to confirm this. The community does not appear to have access to measured data comparing thin and thick client systems in operation in the same situation, allowing direct comparisons to be drawn. This is the main goal of this project. One specific question relates to the overall power use, while it would seem to be obvious that the thin client would require less electricity, what of the server? Two other variations are also considered: it is not uncommon for thin client deployments to continue to use their existing PCs as thin client workstations, with or without modification. Also, attempts by PC makers to reduce the power requirements of their products have given rise to a further variation: the incorporation of low power features in otherwise standard PC technology, working as thick clients. This project was devised to conduct actual measurements in use in a typical university environment. We identified a test area: a mixed administrative and academic office location which supported a range of users, and we made a direct replacement of the current thick client systems with thin client equivalents; in addition, we exchanged a number of PCs operating in thin and thick client mode with devices specifically branded as “low power” PCs and measured their power requirements in both thin and thick modes. We measured the energy consumption at each desktop for the duration of our experiments, and also measured the energy draw of the server designated to supporting the thin client setup, giving us the opportunity to determine the power per user of each technology. Our results show a significant difference in power use between the various candidate technologies, and that a configuration of low power PC in thick client mode returned the lowest power use during our study. We were also aware of other factors surrounding a change such as this: we have addressed the technical issues of implementation and management, and the non-technical or human factors of acceptance and use: all are reported within this document. Finally, our project is necessarily limited to a set of experiments carried out in a particular situation, therefore we use estimation methods to draw wider conclusions and make general observations which should allow others to select appropriate thick or thin client solutions in their situation.

http://www.jisc.ac.uk/whatwedo/programmes/greening...