To provide high availability for services such as mail or bulletin boards, data must be replicated. One way to guarantee consistency of replicated data is to force service operations to occur in the same order at all sites, but this approach is expensive. For some applications a weaker causal operation order can preserve consistency while providing better performance. This paper describes a new way of implementing causal operations. Our technique also supports two other kinds of operations: operations that are totally ordered with respect to one another, and operations that are totally ordered with respect to all other operations. The method performs well in terms of response time, operation processing capacity, amount of stored state, and number and size of messages; it does better than replication methods based on reliable multicast techniques.

This paper addresses the problem of orgamsmg and controllmg actlvltles that mvolve multiple steps of processing and that typically are of long durationWe explore the use of trlggers and transactions to specrfy and organize such long-running actlvltres Triggers offer data-or eventdriven speclficatlon of control flow, and thus provide a flexible and modular framework with which the control structures of the actlvltles can be extended or modified We describe a model based on event-condrtlon-action rules and couphng modes The execution of these rules IS governed by an extended nested transaction model Through a detailed example, we illustrate the utlbty of the various features of the model for chwnmg related steps without sacrificing concurrency, for enforcing Integrity constraints, and for providing flexible failure and exception handlmg

No abstract

This paper describes two techniques that are only loosely related. The first is a method for constructing a highly available service for use in a distributed system. The service presents its clients with a consistent view of its state, but the view may contain old information. Clients can indicate how recent the information must be. The method can be used in any application in which the property of interest is stable: once the property becomes true, it remains true forever.The paper also describes a fault-tolerant garbage collection method for a distributed heap. The method is practical and efficient. Each computer that contains part of the heap does local garbage collection independently, using whatever algorithm it chooses, and without needing to communicate with the other computers that contain parts of the heap. The highly available central service is used to store information about inter.computer references.

No abstract