Abstract

Abstract
Optimistic distributed systems often rely on version vectors or their variants in order to track updates on replicated objects. Some of these mechanisms rely on some form of global configuration or distributed naming protocol in order to assign unique identifiers to each replica. These approaches are incompatible with replica creation under arbitrary partitions, a typical operation mode in mobile or poorly connected environments. Other mechanisms assign unique identifiers relying on statistical correctness. In previous work we have introduced an update tracking mechanism that overcomes these limitations. This paper presents results from recent experimentation, that brought to surface a particular pattern of operation that results in an unforeseen, unlimited growth in space consumption. We also describe informally a new update tracking mechanism that does not exhibit this pathological growth while providing guaranteed unique identifiers for a dynamic number of replicas under arbitrary partitions and the same functionality of version vectors.

Abstract
Causality tracking mechanisms, such as vector clocks and version vectors, rely on mappings from globally unique identifiers to integer counters. In a system with a well known set of entities these ids can be preconfigured and given distinct positions in a vector or distinct names in a mapping. Id management is more problematic in dynamic systems, with large and highly variable number of entities, being worsened when network partitions occur. Present solutions for causality tracking are not appropriate to these increasingly common scenarios. In this paper we introduce Interval Tree Clocks, a novel causality tracking mechanism that can be used in scenarios with a dynamic number of entities, allowing a completely decentralized creation of processes/replicas without need for global identifiers or global coordination. The mechanism has a variable size representation that adapts automatically to the number of existing entities, growing or shrinking appropriately. The representation is so compact that the mechanism can even be considered for scenarios with a fixed number of entities, which makes it a general substitute for vector clocks and version vectors.

Abstract

Abstract

Abstract
Version vectors (VV) are used pervasively to track dependencies between replica versions in multi-version distributed storage systems. In these systems, VV tend to have a dual functionality: identify a version and encode causal dependencies. In this paper, we show that by maintaining the identifier of the version separate from the causal past, it is possible to verify causality in constant time (instead of O(n) for VV) and to precisely track causality with information with size bounded by the degree of replication, and not by the number of concurrent writers. © 2012 Authors.