We consider a DSM system consisting of a number of fail-stop nodes
[9], connected through a communication network.
Each node consists of a processor, a volatile main memory and a non-volatile
secondary memory.
The processors in the system do not share any physical memory and communicate
by message passing.
However, the system can be viewed as a set of processes communicating through
the logically shared memory space, which is composed of the distributed local
memorys of the nodes.
The distributed shared memory space is assumed to consist of a set of
fixed-size pages.
The communication subsystem is assumed to be reliable, however, no assumption
is made on the message delivery order.
Failures considered in the system are transient and a number of concurrent
node failures may happen in the system.
The computation of a process is assumed to be piece-wise deterministic;
that is, the computational states of a process is fully determined by a
sequence of data values provided for the sequence of read operations.
For the memory consistency, the invalidation-based lazy release consistent(LRC)
memory model [3] is assumed, and the model also employs the
multiple-reader, multiple-writer protocol.
A number of different memory semantics for the DSM have been proposed, however,
in this paper, we focus on the LRC memory model for its enhanced performance.
In the LRC memory model, the synchronization operations, such as the acquire, release, barrier, are used to guarantee the correct execution order
between the conflicting operations.
Also, if a process writes on a data page and releases a lock, other processes
can invalidate the old copy of the page and read the new page after it acquires
the corresponding lock.