Fernando de Ferreira Rezende

A lock method for KBMSs using abstraction relationships' semantics

Knowledge Base Management Systems (KBMSs) are a growing research area finding applicability in different domains. As a consequence, the demand for ever-larger knowledge bases (KBs) is growing more and more. Inside this context, knowledge sharing turns out to be a crucial point to be supported by KBMSs. In this paper, we propose a way of controlling knowledge sharing. We show how we obtain serializability of transactions providing many different locking granules, which are based on the semantics of the abstraction relationships. The main benefit of our technique is the high degree of potential concurrency, to be obtained through a logical partitioning of the KB graph and the provision of lock types used for each referenced partition. By this way, we capture more of the semantics contained in a KB graph, through an interpretation of its edges grounded in the abstraction relationships, and make feasible a full exploitation of all inherent parallelism in a knowledge representation approach.

Detection Arcs for Deadlock Management in Nested Transactions and their Performance

In this paper, we address deadlock management in nested transactions. In our strategy, deadlocks are detected through the occurrence of cycles in a waits-for graph (WFG). However, the process of looking for cycles in the WFG considering all its nodes and edges can be very time-consuming. To accelerate this process, we propose the detection arcs. In essence, a detection arc represents a higher level abstraction embodying a hidden waiting relation between two transactions that is caused by a lock wait. Thus, the deadlock detection process needs to traverse only a minimal subset of the WFGs edges when it is started, the set of detection arcs. Therefore, the overall performance of deadlock management is improved, as confirmed by our performance measurements.

Employing Object-Based LSNs in a Recovery Strategy

We present a recovery strategy supporting partial rollbacks of transactions and object-granularity locking, called WALORS (WAL-based andObject-orientedRecoveryStrategy). As the name suggests, WALORS uses the principle of Write-Ahead Logging (WAL). In contrast to other strategies, WALORS stores a Log Sequence Number (LSN) in every object of the database (DB) to correlate the state of the object with logged updates to the object. Due to this feature, WALORS does not need to employ a repeating history (redo all) paradigm. Instead, it supports selective undo as well as selective redo passes. To avoid the problems of supporting fine-granularity locking in the context of WAL, WALORS employs special control structures which enable it not having to write compensation log records (CLRs) from CLRs and guaranteeing its idempotence even in the face of repeated failures or of nested rollbacks. WALORS does operation logging of all updates, including the ones performed during rollbacks, works with fuzzy checkpoints, supports media recovery, and is flexible enough w.r.t. the kinds of buffer management policies being implemented.