Dodo: A User-level System for Exploiting Idle Memory in WorkstationClusters

In this paper, we present the design and implementation of Dodo, an efficientuser-level system for harvesting idle memory in off-the-shelf clusters ofworkstations. Dodo enables data-intensive applications to use remote memory ina cluster as an intermediate cache between local memory and disk. It requiresno modifications to the operating system and/or processor firmware and is henceportable to multiple platforms. Further, the memory recruitment policy used byDodo is designed to minimize any delays experienced by the owner of desktopmachines whose memory is harvested by Dodo.Our implementation of Dodo is operational and currently runs on Linux 2.0.35.For communication, Dodo can use either UDP/IP or U-Net, the low-latencyuser-level network architecture developed by von Eicken et al. We evaluatedthe performance improvements that can be achieved by using Dodo for two realapplications and three synthetic benchmarks. Our results show that speedupsobtained for an application are highly dependent on its I/O access pattern anddata set sizes. Significant speedups (between 2 and 3) were obtained forapplications whose working sets are larger than the local memory on aworkstation but smaller than aggregate memory available on the cluster and forapplications that can benefit from the zero-seek nature of remote memory.