Wecome to DCPerf 2011

The first international workshop on Data Center Performance

Data centers are the backbone infrastructure for tomorrow's information technology. Their advantages are efficient resource provisioning and low operational costs for supporting a wide range of computing, be it in business, scientific or mobile/pervasive environments. Owing to the rapid growth in user-defined and user-generated programs, applications and files, the range of services provided at data centers will expand tremendously and unpredictably. The high volume of mixed workloads and the diversity of services offered render the performance optimization of data centers ever more challenging. Moreover, important optimization criteria, such as scalability, reliability, manageability, power efficiency, area density, operating costs and many more, are often mutually exclusive to some extent. On top of that, the increasing mobility of users across geographically distributed areas adds another dimension to optimizing data center performance.

To meet the wide range of quality-of-service requirements, for example in terms of throughput, latency and security, performance optimization at data centers is typically adopted in the areas of hardware components, server systems, and software.  The reliability of a data center in hosting services and storing data chiefly hinges on the seamless provisioning of computing and communication resources, such as cores, memory, I/O and networks, as well as facility resources such as power supplies.  State-of-the-art visualization and management tools are employed on different layers of the server system so that the hypervisor can efficiently multiplex the corresponding stack of hardware, operating system, middleware, and applications to a large number of users, who potentially are highly mobile across a large geographical area. Moreover, a stack of multiple virtualized layers will tremendously increase the difficulty of optimizing data-center performance, in particular for diverse workloads. As a result, adaptive and autonomic optimization strategies are essential in managing the multi-dimensional complexity of data center performance, especially in an online fashion.

The goal of this workshop is to promote a community-wide discussion to find and identify suitable strategies to enable effective and scalable data center performance optimization. We are looking for papers that present new techniques, introduce new methodologies, propose new research directions, or discuss strategies for resolving open performance problems at all layers of a data center. The focus will be on the data center's system-level aspects, communication, virtualization and performance optimization.