Exception handling and fault-tolerance in multimedia synchronization

This paper contributes toward exception handling and fault-tolerance in multimedia presentation. Our study is based on the well-known four phases in fault-tolerant computing: fault detection, damage assessment, error recovery, and continued service. We define a fault in multimedia synchronization and a simple fault detection mechanism. Using the concept of a partner set of a media stream, we assess the damage caused to media presentation due to a fault. From the point of error recovery and reducing synchronization failures, we introduce the ideas of a k-cycle virtual fault. A k-cycle virtual fault suggests the possibility of a failure in the future after k presentation cycles. Detection of a possible presentation failure in the future gives lead time to take corrective measures to avoid the failure. In order to handle exception conditions during a synchronization failure, we define pinned and sliding semantics of media presentation. These two semantics allow us to define different levels of quality of presentation during a failure. Finally, we present the detailed design of a fault-tolerant presentation architecture and prove its properties. We discuss how the ideas of virtual fault and damage assessment can be used in generating useful information for the underlying data transfer protocol so that synchronization failures can be reduced.