Ana Nunes

Distributed transaction processing has benefited greatly from optimistic concurrency control protocols thus avoiding costly fine-grained synchronization. However, the performance of these protocols degrades significantly when the workload increases, namely, by leading to a substantial amount of aborted transactions due to concurrency conflicts. Our approach stems from the observation that when the abort rate increases with the load as already executed transactions queue for longer periods of time waiting for their turn to be certified and committed. We thus propose an adaptive algorithm for judiciously scheduling transactions to minimize the time during which these are vulnerable to being aborted by concurrent transactions, thereby reducing the overall abort rate. We do so by throttling transaction execution using an adaptive mechanism based on the locally known state of globally executing transactions, that includes out-of-order execution.
Our evaluation using traces from the industry standard TPC-E workload shows that the amount of aborted transactions can be kept bounded as system load increases, while at the same time fully utilizing system resources and thus scaling transaction processing throughput.

Publish/subscribe mechanisms for scalable event dissemination are a core component of many distributed systems ranging from EAI middleware to news dissemination in the Internet. Hence, a lot of research has been done on overlay networks for efficient decentralized topic-based routing. Specifically, in gossip-based dissemination, approximating nodes with shared interests in the overlay makes dissemination more efficient. Unfortunately, this usually requires fully disclosing interests to nearby nodes and impacts reliability due to clustering.
In this paper we address this by starting with multiple overlays, one for each topic subscribed, that then separately self-organize to share a large number of physical connections, thereby leading to reduced message traffic and maintenance overhead. This is achieved without a node ever disclosing an interest to another node that doesn’t share it and without impacting the robustness of the overlay. Besides presenting the overlay maintenance protocol, we evaluate it using simulation in order to validate our results.

Althought optimistic concurrency control protocols have increasingly been used in distributed database management systems, they imply a trade-of between the number of transactions that can be executed concurrently, hence, the peak throughput, and transactions aborted due to conflicts.
We propose a novel optimistic concurrency control mechanism that controls transaction abort rate by minimizing the time during which transactions are vulnerable to abort, without compromising throughput. Briefly, we throttle transaction execution with an adaptive mechanism based on the state of the transaction queues while allowing out-of-order execution based on expected transaction latency. Prelimi- nary evaluation shows that this provides a substantial improvement in committed transaction throughput.