CS Lunch talk Monday, October 3
Dan Amato '07 and Alexandr Pshenichkin '07
will discuss their summer research work
with Professor Newhall
"
Reliability for Nswap
"
A cluster is a collection of interconnected computers (workstations using
off-the-shelf hardware) that present a single system image of a large parallel
computer. Nswap is a project that implements a shared memory system, allowing
individual machines in a cluster to use other nodes' free RAM as a swap space;
this is significantly faster than swapping to disk. When individual nodes in
the cluster fail, however, processes on other nodes are compromised: a system
for recovering data swapped out to memory on failed nodes is needed. This talk
will present a general overview of Nswap and discuss the reliability system we
analyzed and developed this summer.