Reducing host load, network load, and latency in a distributed shared memory | IEEE Conference Publication | IEEE Xplore