In this chapter, we focus on the issue of network performance in volunteer networks. For some applications, communication performance plays a nontrivial role in the overall performance. It is this class of applications that motivates our work. Network performance can become important when peers or volunteers are interacting with each other and/or with other nodes, for example, servers or external nodes. We focus on both techniques to estimate network performance efficiently and how to use such performance in a variety of volunteer computing scenarios. We first consider how time-bounded execution of data-intensive services can be achieved across volunteers using a form of communication makespan estimation relying on application-specific communication measurements. This problem reduces to a server selection problem where the data is replicated across a set of servers. Next, we consider how second-hand communication information can be estimated and propagated to peers through the OPEN framework. Finally, we describe a tool for improving communications in volunteer networks, the network dashboard. This tool can identify superior application-level network routing paths in volunteer networks based on a variety of different metrics. We present results that indicate that superior paths in terms of TCP/UDP bandwidth, delay, and jitter can be found.
|Original language||English (US)|
|Title of host publication||Desktop Grid Computing|
|Number of pages||24|
|State||Published - Jan 1 2012|