Quality of Service

Transferring large quantities of data is a frequently invoked process in many scientific applications, and these transfers have high quality requirements on the infrastructure services. The Quality of Service (QoS) related research in SNE focuses on the optimization of the service quality that infrastructures deliver to high level applications.

The recent emergence of advanced network infrastructures for e-Science enables tuning of network performance at the application level. The idea of providing data intensive applications with deterministic point-to-point connections was fostered by a community of research networks, later organized in the Global Lambda Integrated Facility (GLIF). A lambda is a light wavelength used to carry information in optical systems. This community provides a global network to support data-intensive scientific research, and also supports middleware development for optical networking. The ideas in this community led to the concept of hybrid networking, the offering of packet switched (IP) services and circuit switched connections over the same physical network infrastructure. In the meantime, the available speeds of lightpaths keep growing. A lightpath is a network circuit transmitting data between two end points using a specific wavelength. While 10Gbit/sec links were introduced only a few years ago, 40Gbit/sec and 100Gbit/sec links are now becoming available to application developers. These kinds of links provide unique opportunities for transport of high-quality media. By including network resources in the scheduling loop, the high level application gets an extra opportunity to optimize execution and improve performance.

These new opportunities come with challenges. Not all network infrastructures provide the network services for reserving specific connections or allocating network bandwidth; the service invocation in different network domains is often proprietary and not easily extensible, and makes a request for network service provisioning across sites difficult. Scheduling network resources requires knowledge on the current state of the network, which implies the existence of a sophisticated monitoring system. The solutions to these issues require not only the integration among information sources from distributed infrastructure domains, but also the intelligence to invoke different levels of network and application control services.

By enhancing the awareness between infrastructure delivery and the programming and execution model of data intensive applications, we aim at systematic solutions to modeling, programming, and solving the constraints between QoS attributes at different service layers in the physical/virtual infrastructures.

Comments are closed.