Donald E. Young

1000 islands: an integrated approach to resource management for virtualized data centers

Recent advances in hardware and software virtualization offer unprecedented management capabilities for the mapping of virtual resources to physical resources. It is highly desirable to further create a “service hosting abstraction” that allows application owners to focus on service level objectives (SLOs) for their applications. This calls for a resource management solution that achieves the SLOs for many applications in response to changing data center conditions and hides the complexity from both application owners and data center operators. In this paper, we describe an automated capacity and workload management system that integrates multiple resource controllers at three different scopes and time scales. Simulation and experimental results confirm that such an integrated solution ensures efficient and effective use of data center resources while reducing service level violations for high priority applications.

Towards a universal marketplace over the web: statistical multi-label classification of service provider forms with simulated annealing

There is a growing number of service providers that a consumer can interact with over the web to learn their service terms. The service terms, such as price and time to completion of the service, depend on the consumers particular specifications. For instance, a printing services provider would need from its customers specifications such as the size of paper, type of ink, proofing and perforation. In a few sectors, there exist marketplace sites that provide consumers with specifications forms, which the consumer can fill out to learn the service terms of multiple service providers. Unfortunately, there are only a few such marketplace sites, and they cover a few sectors. At HP Labs, we are working towards building a universal marketplace site, i.e., a marketplace site that covers thousands of sectors and hundreds of providers per sector. One issue in this domain is the automated discovery/retrieval of the specifications for each sector. We address it through extracting and analyzing content from the websites of the service providers listed in business directories. The challenge is that each service provider is often listed under multiple service categories in a business directory, making it infeasible to utilize standard supervised learning techniques. We address this challenge through employing a multilabel statistical clustering approach within an expectation-maximization framework. We implement our solution to retrieve specifications for 3000 sectors, representing more than 300,000 service providers. We discuss our results within the context of the services needed to design a marketing campaign for a small business.

Service Selection in Business Service Ecosystem

A world-wide community of service providers has a presence on the web, and people seeking services typically go to the web as an initial place to search for them. Service selection is comprised of two steps: finding service candidates using search engines and selecting those which meet desired service properties best. Within the context of Web Services, the service selection problem has been solved through common description frameworks that make use of ontologies and service registries. However, the majority of service providers on the web does not use such frameworks and rather make service descriptions available on their web sites that provide human targeted content. This paper addresses the service selection problem under the assumption that a common service description framework does not exist, and services have to be selected using the more unstructured information available on the web. The approach described in this paper has the following steps. Search engines are employed to find service candidates from dense requirement formulations extracted from user input. Text classification techniques are used to identify services and service properties from web content retrieved from search links. Service candidates are then ranked based on how well they support desired properties. Initial experiments have been conducted to validate the approach.