Abstract : High-performance cluster networks achieve very high throughput thanks to zero-copy techniques that require pinning of application buffers in physical memory. The Open-MX stack implements message passing over generic Ethernet hardware with similar needs. We present the design of an innovative pinning model in Open-MX based on the decoupling of memory pinning from the application. This idea eases the implementation of a reliable pinning cache in the kernel and enables full overlap of pinning with communication. Performance evaluation shows that both these optimizations bring interesting throughput improvements.
Document type :
IEEE. Workshop on Communication Architecture for Clusters, held in conjunction with IPDPS 2009, May 2009, Rome, Italy. <10.1109/IPDPS.2009.5160888>
Brice Goglin. Decoupling Memory Pinning from the Application with Overlapped on-Demand Pinning and MMU Notifiers. IEEE. Workshop on Communication Architecture for Clusters, held in conjunction with IPDPS 2009, May 2009, Rome, Italy. <10.1109/IPDPS.2009.5160888>. <inria-00356236>