A Utility Data Center (UDC) provides a flexible, cost-effective solution by using advanced management software to automatically reassign resources in response to changing business and IT requirements. The UDC architecture is an ideal platform to support the efficient hosting of applications for Internet services. Most Internet applications are difficult to manage due to their highly variable service demand. Typically, current service providers significantly over provision the amount of resources needed to support such applications by tailoring the resources to meet maximum expected load. The UDC infrastructure provides a set of new management capabilities for requesting/releasing the system resources to dynamically provision the application demands and their requirements.

The area of multimedia services in a networked environment is one of the rapidly expanding fields in today's technological world. The delivery of continuous media from a central server complex to a large number of (geographically distributed) clients is a challenging and resource intensive task.

Our ultimate goal is to design and implement a "utility-aware" streaming media service that automatically requests/releases the necessary resources from the UDC infrastructure to adapt to variable workload demand.

Among the prerequisites to a design of such a utility service is the ability to measure the capacity of media servers (nominal and currently in use) in order to evaluate the amount of available system resources for admitting new client requests or to request/release resources as needed.

We are working on a new unified framework for

• measuring a media service capacity via a set of basic benchmarks;
• deriving the resource requirements (a cost) of a particular stream from the basic benchmarks measurements;
• estimating a currently available/used service capacity under the realistic workloads via a new workload-aware performance model based on a cost function.

Understanding the nature of application workloads is crucial to properly designing and provisioning current and future services. The synthetic workload generator is a facility for generating realistic and reproducible synthetic workloads for use in performance experiments aiming to design new effective streaming media delivery architectures and strategies. For such experiments, the generated workload must not only mimic the highly dynamic resource-utilization patterns found on today's media systems but also provide flexible means to generate more intensive, bursty and diverse workloads for future media systems.

We are working on a synthetic workload generator for streaming media applications (called MediSyn ), which we plan to release as an open source code to researchers of industrial Labs and academia.

We plan to use the prototype of our synthetic workload generator in performance experiments for design of streaming media utility service in Utility Data Center environment.

Related Papers and Reports

• L. Cherkasova, W. Tang, S. Singhal: An SLA-Oriented Capacity Planning Tool for Streaming Media Services. Proc. of the International Conference on Dependable Systems and Networks, (DSN'2004), Florence,Italy, June 28-July 1, 2004.

• L. Cherkasova, W. Tang: Sizing the Streaming Media Cluster Solution for a Given Workload In Proc. of the 4th IEEE/ACM International Symposium on Cluster Computing and the Grid (CCGrid'2004), Chicago, USA, April 19-22, 2004.
  

• L. Cherkasova, W. Tang: Providing Resource Allocation and Performance Isolation in a Shared Streaming-Media Hosting Service. In Proc. of ACM Symposium on Applied Computing (SAC'2004), March, 2004.

• L. Cherkasova, W. Tang. Capacity Planning Tool for Streaming Media Services In Proceedings of ACM Multimedia'2003, Berkeley, CA, November 2-8, 2003.

• L. Cherkasova, L. Staley. Building a Performance Model of Streaming Media Applications in a Utility Data Center Environment. Proceedings of the 3rd IEEE/ACM International Symposium on Cluster Computing and the Grid (CCGrid'2003), Tokyo, Japan, May 12-15, 2003

• L.Cherkasova, J. Lee. FastReplica: Efficient Large File Distribution within Content Delivery Networks. Proceedings of the 4th USENIX Symposium on Internet Technologies and Systems (USITS'2003), Seattle, Washington, March 26-28, 2003.

• W. Tang, Y. Fu, L. Cherkasova, A. Vahdat. Long-term Streaming Media Server Workload Analysis and Modeling. HP Laboratories Report No. HPL-2003-23, 2003.

• L. Cherkasova, L. Staley. Measuring the Capacity of a Streaming Media Server in a Utility Data Center Environment. Proceedings of ACM Multimedia'2002, Juan Les Pines, France, Dec. 1-6, 2002.