The problem of system failure and recovery of distributed shared virtual memory (DSVM) is studied. Most DSVM systems use the notion of tokens to indicate a site's access rights on the data pages it caches, and a locating scheme to get to the most up-to-date version of a data page. The problem is to recovery this token directory after a site has failed. The authors' solution is to treat the token directory at each site as a fragment of a global token database and the page migration activities as token transactions that update this distributed database. By the use of the unilateral commit protocol for token transactions, fast recovery of the token state at minimal run-time overhead of token transaction execution is achieved.>

A distributed memory hierarchy (DMH) is a memory system consisting of storage modules distributed over a high-bandwidth local area network. It provides for transaction applications an abstraction of single virtual memory space to which shared data are mapped. As in a conventional memory hierarchy (MH) in a single-machine system, a DMH is responsible for locating, migrating, and caching data pages; however, unlike a conventional MH, a DMH must do so across the storage modules in a network. In addition, a DMH must handle the problem of propagation of transaction updates preserving serializability of transactions. The performance of a DMH system is strongly influenced by concurrency control and update propagation. It is also crucial that performance analysis accounts for memory resources and network requirements. A DMH system is presented, the tradeoffs between conservative and aggressive update propagation strategies are defined, and promising new strategies are identified.>

Executing a transaction in a conventional distributed database system involves the execution of several subtransactions, each at a remote site where the data reside and running a two-phase commit protocol at the end of the transaction. With the advent of fast communication networks, we consider an alternative paradigm where the remote data being accessed are dynamically migrated to the initiation site of the transaction. One example of such a system is a distributed shared virtual memory system.

Synchronization in DSVM (distributed shared virtual memory) can be approached top-down by first understanding the synchronization needs at the process level instead of only at the memory access level. The authors demonstrate this idea in the context of transaction synchronization, devising two-phase locking-based algorithms under two DSVM scenarios: with and without an underlying memory coherence system. They compare the performances of the two algorithms and argue that significant performance gain can potentially result from bypassing memory coherence and supporting process synchronization directly on distributed memory. They also study the role of the optimistic algorithms in transaction synchronization in DSVM and show that some optimistic policy appears promising under the scenarios studied.>