Developing a QoS-based Tasklet trading system

Tasklets provide means for lightweight distributed computing. Based on virtual machines, they allow to utilize excess capacity in cloud computing as well as integrate cloud resources into mobile computing. So far, the Tasklets architecture only provides a best-effort computing abstraction while using direct addressing of peers. However, a higher level of service is often desirable, as is found in other technologies for distributed computing. For example, in the Internet protocol suite, the Transport layer provides the choice of reliable or unreliable communication on top of the best-effort Internet layer. In middle-ware technology, distribution transparency masks where a specific service is located. Similar guarantees and abstractions in other systems are achieved via Quality of Service (QoS) parameters and mechanisms. In this paper, we discuss QoS categories for Tasklets, develop the necessary system extensions for QoS integration, and present a QoS-based Tasklet trading system. We evaluate our approach in a prototype system by trading Tasklets of two different prioritization levels.