Shared Virtual Memory: Progress and Challenges
When the idea of implementing a shared address space
in software at page granularity across a network of computers---i.e.
the shared virtual memory or SVM approach---was first
introduced, few would have predicted that
a decade would not be enough to exhaust its research
potential or even understand its performance capabilities.
Software shared memory became a very active
research area, with many groups designing new
protocols, consistency models and systems.
These groups inspired and motivated one another, building on
one another's ideas to push performance higher.
This paper places the multi-track flow of ideas and
results so far in a framework, with the goal of identifying
the lessons learned so far and the major remaining challenges.