Publications

Partial replication in the Database State Machine

This paper investigates the use of partial replication in the Database State Machine approach introduced earlier for fully replicated databases. It builds on the order and atomicity properties of group communication primitives to achieve strong consistency and proposes two new abstractions: Resilient Atomic Commit and Fast Atomic Broadcast. Even with atomic broadcast, partial replication requires a termination protocol such as atomic commit to ensure transaction atomicity. With Resilient Atomic Commit our termination protocol allows the commit of a transaction despite the failure of some of the participants. Preliminary performance studies suggest that the additional cost of supporting partial replication can be mitigated through the use of Fast Atomic Broadcast.

Optimistic Total Order in Wide Area Networks

Total order multicast greatly simplifies the implementation of fault-tolerant services using the replicated state machine approach. The additional latency of total ordering can be masked by taking advantage of spontaneous ordering observed in LANs: A tentative delivery allows the application to proceed in parallel with the ordering protocol. The effectiveness of the technique rests on the optimistic assumption that a large share of correctly ordered tentative deliveries offsets the cost of undoing the effect of mistakes. This paper proposes a simple technique which enables the usage of optimistic delivery also in WANs with much larger transmission delays where the optimistic assumption does not normally hold. Our proposal exploits local clocks and the stability of network delays to reduce the mistakes in the ordering of tentative deliveries. An experimental evaluation of a modified sequencer-based protocol is presented, illustrating the usefulness of the approach in fault-tolerant database management.

Avaliação de um SGBD replicado usando simulaçãao de redes

A replicação de sistemas de gestão de bases de dados (SGBD) é um mecanismo fundamental para a fiabilidade de sistemas de informação. Em sistemas geograficamente distribuídos é ainda fundamental para recuperação de desastres e disponibilidade ubíqua de dados. Uma técnica de replicação recentemente proposta é a Database State Machine (DBSM), que promete aliar fiabilidade a elevado desempenho tirando partido de sistemas de comunicação em grupo. A avaliação do desempenho desta técnica tem no entanto sido efectuada com redes de comunicação demasiado simples ou irrealistas e com uma carga não representativa. Este artigo propõe uma avaliação rigorosa de uma concretização desta técnica de replicação, aliando um modelo de simulação realista de redes de comunicação com uma geração de carga efectuada de acordo com os padrões das medidas de desempenho elaboradas pelo Transaction Processing Council (TPC). Os resultados obtidos confirmam o interesse desta técnica em redes locais mas mostram que o seu desempenho é condicionado pelas características da rede e da carga.

Evaluating database replication in ESCADA (position paper)

 

Evaluating the performance of the database state machine

Replication of database servers using the Database State Machine (DBSM) approach has recently been the subject of much attention as it promises both good performance and increased reliability. Ful lling the promise of performance, however, requires that the impact of environment parameters as well as of design and implementation decisions are evaluated in a realistic setting which allows experimentation with con guration and environment parameters. This paper introduces a model that combines simulated network and database engine components with real implementations of termination and communication protocols. This allows us to precisely evaluate the protocols' performance when subjected to a realistic load generated by the simulated database engine. It also allows us to evaluate the impact of the protocol overhead on the overall performance of the database system in several environments. Besides describing the design and validation of the simulation model, results obtained with prototype implementations of the protocols are presented.

Evaluating certification protocols in the partial database state machine

Partial replication is an alluring technique to ensure the reliability of very large and geographically distributed databases while, at the same time, offering good performance. Partial replication is done by splitting the database according to application semantics and then by replicating each fragment at a subset of the available sites. Access locality should allow that each transaction needs only a small subset of all sites to execute and commit thus reducing processing and communication overhead associated with replication. The advantages of partial replication have however to be weighted against the added complexity that is required to manage it. In fact, if the chosen configuration cannot make transactions execute locally or if the overhead of consistency protocols offsets the savings of locality, potential gains cannot be realized. Unfortunately, both these issues are heavily dependent on the application used for benchmarking thus rendering simplistic benchmarks useless. In this paper we present a detailed analysis of Partial Database State Machine~(PDBSM) replication by comparing alternative partial replication protocols with full replication. This is done using a realistic scenario based on a detailed network simulator and access patterns from an industry standard database benchmark. The results obtained allow us to identify the best configuration for typical on-line transaction processing applications.