As raw system and network performance continues to improve at exponential rates, the utility of many services is increasingly limited by availability rather than performance. A key approach to improving availability involves replicating the service across multiple, wide-area sites. However, replication introduces well-known tradeoffs between service consistency and availability. Thus, this paper explores the benefits of dynamically trading consistency for availability using a continuous consistency model. In this model, applications specify a maximum deviation from strong consistency on a per-replica basis. In this paper, we: i) evaluate availability of a prototype replication system running across the Internet as a function of consistency level, consistency protocol, and failure characteristics, ii) demonstrate that simple optimizations to existing consistency protocols result in significant availability improvements (more than an order of magnitude in some scenarios), iii) use our experience with these optimizations to prove tight upper bounds on the availability of services, and iv) show that maximizing availability typically entails remaining as close to strong consistency as possible during times of good connectivity, resulting in a communication versus availability tradeoff.

Abstract. IceCube is a system for optimistic replication, supporting collaborative work and mobile computing. It lets users write to shared data with no mutual synchronisation; however replicas diverge and must be reconciled. IceCube is a general-purpose reconciliation engine, parameterised by “constraints ” capturing data semantics and user intents. IceCube combines logs of disconnected actions into near-optimal reconciliation schedules that honour the constraints. IceCube features a simple, high-level, systematic API. It seamlessly integrates diverse applications, sharing various data, and run by concurrent users. This paper focus on the IceCube API and algorithms. Application experience indicates that IceCube simplifies application design, supports a wide variety of application semantics, and seamlessly integrates diverse applications. On a realistic benchmark, IceCube runs at reasonable speeds and scales to large input sets. 1

In a distributed system, optimistic replication enables users on different sites to query and update their local replicas of shared databases without a priori synchronization. Replicas may diverge, and updates must be reconciled; reconciliation is a difficult problem in the presence of conflicts, alternative execution paths, and dependencies between transactions. We present SqlIce-Cube, a general-purpose reconciliation system. SqlIceCube automatically extracts significant semantic relations from the program text of transactions. Examples of relations are a transaction that depends on another, helps another, hinders another, constitutes an alternative to another, etc. In turn these semantic relations feed into the SqlIceCube scheduler, which generates and executes a combination of transactions with the highest possible value. The approach is general and supports combinations of transactions across a variety of different applications. Application experience and benchmarks show the viability of the approach. SqlIceCube correctly extracts semantic relations from an interesting variety of application transactions, performs well, and scales well to large input sizes. 1

Managing distributed state is a difficult challenge for building scalable, distributed, wide-area applications. This project presents the design of an infrastructure, called Ivory, to simplify construction of distributed applications by automatically caching and replicating data structures and code. We illustrate the use of our infrastructure in service caches that replicate Web service code and data used to generate dynamic content. The service cache relies on Ivory to maintain consistency of cached data as a basis for scalable dynamic Web services.

Wireless devices currently provide real-time access to personalized information such as headlines, email, stock quotes, and online auctions. Retrieving all updates to such rapidly changing information is wasteful of both network bandwidth and battery power. Existing consistency models for such services allow for only coarse-grained timeouts on how often information should be retrieved. In this paper, we argue for the benefits of a continuous consistency model for user access to Internet portal services. Using this model, users are able to specify the maximum error in their view of the data. For instance, users may specify that they wish to receive an updated stock quote only if the users view diverges from the actual value by more than 3%. Services may also use application-specific semantics to control data consistency— e.g., new bids carry more weight as the auction draws to a close. We use a simulator to model the bandwidth and energy benefits available from a more flexible consistency model. Such benefits depend upon the rate at which underlying data values change. To capture representative distributions, we use a trace-based study of updates to weather, news, and stock quotes from a popular portal to determine representative distribution ranges. Our initial results indicate bandwidth and energy savings that increase as users are willing to tolerate larger bounds on data accuracy. 1.