Distributed applications relying on replication of data objects in several sites are becoming popular. One key to their success is that they rely on reliable broadcast/multicast protocols. Two classes of approaches to build reliable broadcast services are considered: clock-driven and the clockless. The former relies on the existence of a global timebase, whereas the latter does not. Clock-driven and clockless protocols have often been classified as equivalent to synchronous and asynchronous, respectively. The clock as an implementation tool is, however, not mandatory to achieve synchronism. By measuring the quality of synchronism by steadiness and tightness--a metrics that applies equally to both classes--it is shown that the criteria of suitability for real-time and fault-tolerance can be extended to clockless protocols.
|Original language||English (US)|
|Title of host publication||Proceedings - Workshop on the Management of Replicated Data|
|Publisher||Publ by IEEEPiscataway, NJ, United States|
|Number of pages||5|
|State||Published - Dec 1 1990|