Self-tuned remote execution for pervasive computing：（对普适计算Self-tuned远程执行）.pdfVIP

Self-Tuned Remote Execution for Pervasive Computing Jason Flinn, Dushyanth Narayanan and M. Satyanarayanan School of Computer Science Carnegie Mellon University Abstract When locating functionality, Spectra must balance the tra- ditional goal of minimizing application latency with new Pervasive computing creates environments saturated goals such as maximizing battery lifetime. It must allow with computing and communication capability, yet grace- for wider variation in resources such as CPU and network fully integrated with human users. Remote execution has bandwidth and monitor new resources such as energy use a natural role to play in such environments, since it lets and cache state. applications simultaneously leverage the mobility of small Pervasiveness causes additional complexity, and it is un- devices and the greater resources of large devices. In this reasonable to leave the burden of handling this complex- paper, we describe Spectra, a remote execution system de- ity to applications. Spectra does not require applications to signedfor pervasive environments. specify resource requirements for a variety of platforms and Spectra monitors resources such as battery energy and output qualities. Instead, it is self-tuning—it monitors ap- file cache state which are especially important for mo- plication resource usage in order to predict future behavior. bile clients. It also dynamically balances energy use and quality goals with traditional performance concerns to de- 2 Design consideratio