Computer communications networks are becoming both more complex and more interconnected.
Applications such as e-mail, podcasting and telephone calls are all expected to work seamlessly in diverse and heterogeneous network environments and traverse systems managed by different entities. Watching a video clip posted at a major news Web site from a home computer, for example, might involve the bits crossing five independently run networks.
Making such an experience seamless requires addressing a number of major technical challenges. Among them:
How can service providers ensure good service quality -- so that customers' videos play smoothly and phone calls aren’t dropped -- given that no one provider controls the end-to-end path traversed by the application?
How does a network provider maintain data security as it travels across networks?
How do providers manage their networks and detect/diagnose problems in real-time when their services are so dependent on other, independently run networks?
Building on HP's expertise, HP Labs is solving these problems by focusing on inferential network measurement and control.
Our research is designed to allow providers and users to infer the status of the broader network by measuring those key parts of the interconnected network that can be accessed. A provider can then use mechanisms over which it has control to make its part of the network run better automatically and in real time.
Research projects in this area include:
Scalable sensing service -- taking a holistic approach to sensing data from all readable components of a network, including nodes, servers, hubs, routers and caches. The scalability of this work is being tested on the PlanetLab global-scale network test bed.
Chart sensing infrastructure -- a project funded by the U.S. Defense Advanced Research Projects Agency (DARPA) to make the TCP Internet protocol to throughputs higher by a factor of 10, given a variety of link impairments. This is also being tested and demonstrated on PlanetLab.
Home network management -- using inferential sensing and management to inform consumers about network status, help them diagnose networking problems, and allow them to better manage home networks.
Mesh networking management -- providing sensing and management tools for wireless ‘mesh’ networks in which multiple local wireless nodes are linked together to form a multi-hop wireless network.
Service-oriented architectures -- working with HP Software on service-oriented architecture (SOA) management. By better compartmentalizing service application software -- and creating new tools to better manage the compartmentalization process -- SOAs can be more quickly leveraged to develop new, more complicated services.