Pablo Galdámez

Measuring Availability in Optimistic Partition-Tolerant Systems with Data Constraints

Replicated systems that run over partitionable environments, can exhibit increased availability if isolated partitions are allowed to optimistically continue their execution independently. This availability gain is traded against consistency, since several replicas of the same objects could be updated separately. Once partitioning terminates, divergences in the replicated state needs to be reconciled. One way to reconcile the state consists of letting the application manually solve inconsistencies. However, there are several situations where automatic reconciliation of the replicated state is meaningful. We have implemented replication and automatic reconciliation protocols that can be used as building blocks in a partition-tolerant middleware. The novelty of the protocols is the continuous service of the application even during the reconciliation process. A prototype system is experimentally evaluated to illustrate the increased availability despite network partitions.

Implementing Network Partition-Aware Fault-Tolerant CORBA Systems

The current standard for fault-tolerance in the Common Object Request Broker Architecture (CORBA) does not support network partitioning. However, distributed systems, and those deployed on wide area networks in particular, are susceptible to network partitions. The contribution of this paper is the description of the design and implementation of a CORBA fault-tolerance add-on for partitionable environments. Our solution can be applied to an off-the-shelf Object Request Broker, without having access to the ORBs source code and with minimal changes to existing CORBA applications. The system distinguishes itself from existing solutions in the way different replication and reconciliation strategies can be implemented easily. Furthermore, we provide a novel replication and reconciliation protocol that increases the availability of systems, by allowing operations in all partitions to continue

CORBA Replication Support for Fault-Tolerance in a Partitionable Distributed System

The common object request broker architecture (CORBA) specification originally did not include any support for fault-tolerance. The fault-tolerant CORBA standard was added to address this issue. One drawback of the standard is that it does not include fault-tolerance in the case of network partitioning faults. The main contribution of this paper is the design of a fault-tolerance CORBA add-on for partitionable environments. In contrast to other solutions, our modular design separates replication and reconciliation policies from the basic replication mechanisms. This modularity allows the replication and reconciliation strategies to be modified easily

Flexible Management of Consistency and Availability of Networked Data Replications

We describe a family of three replication protocols, each of which can operate in three different modes of consistency. The protocols are tailored to satisfy the availability demands of interconnected databases that have a high degree of data locality. The protocols accomplish a grade of transaction completion which does not compromise availability, and ensure the consistency of replicas also if a transaction needs to be aborted. Flexibility of query answering is understood as optimizing the tradeoff between consistency and availability, i.e., between correctness and timeliness of query answering. This is achieved by choosing an appropriate protocol alternative, and changing the consistency mode of operation during the session, as appropriate for a given transaction.

ROI: an invocation mechanism for replicated objects

The reliable object invocation mechanism provided by HIDRA for the coordinator-cohort and the passive replication models offers support to ensure that all the replicas of the object being invoked are correctly updated before such an invocation is terminated. This mechanism also ensures that if a primary or coordinator replica crashes, the client is able to reconnect to the previously initiated invocations, collecting their results without requiring their reexecution. All this support is provided transparently to the client of the replicated objects, which does not notice any difference in respect to the invocations made to non-replicated objects. Moreover, the protocols described in the paper deal also with the failure of any of the objects involved in this kind of invocations.

DeDiSys Lite: an environment for evaluating replication protocols in partitionable distributed object systems

Distributed object systems for partitionable systems present a challenge, in that there is a trade-off between availability and consistency. Changes in one partition are not visible in another partition. Therefore, if strong consistency is required, certain operations cannot be permitted. This reduces availability. In the DeDiSys project we aim at allowing this trade-off between consistency and availability to be configurable. The DeDiSys distributed object system relies heavily on replication protocols that allow high-availability, whilst ensuring a level of consistency that is required by a particular application. We have developed DeDiSys Lite, a prototype of the DeDiSys system, which provides a platform to implement and evaluate these replication protocols. Infrastructure components are provided in a minimal implementation. Configuration files allow system parameters, such as the degree of replication or the nesting of object invocations, to be modified, without having to adapt application code. We use DeDiSys Lite as both a simulation environment for the development of new replication protocols and as a basis for the continuous development of the DeDiSys system. Some results obtained using the platform to optimise a new replication protocol are presented in this paper.

Enhancing the availability of networked database services by replication and consistency maintenance

We describe a middleware platform for maintaining the consistency of replicated data called COPla (Common Object Platform). The purpose of replication is to enhance the availability of data and services in distributed database networks. That is orthogonal to recovery strategies of backed-up snapshots, logs and other measures to alleviate database downtimes. A range of different consistency modes ensures the correctness of replicated data.

HMM: A Cluster Membership Service

The Hidra Membership Monitor (HMM) is a distributed service that maintains the current set of active nodes in a cluster of machines. This protocol allows the detection of multiple machine joins or failures in a unique reconfiguration, using a low amount of messages (with a cost that is linear on the number of nodes). These membership services are needed to detect cluster changes as soon as possible, initiating then the reconfiguration of the cluster state, where support for replicated objects has been included. The HMM also manages and synchronises the reconfiguration steps needed by the kernel and Hidra components of each node, ensuring that all of them take the same steps at once. Thus, our system does not need an atomic multicast protocol to deliver the messages in these reconfiguration steps. All these services provide the basis to develop reliable intracluster transport protocols and to reduce the reconfiguration time of replicated objects and services.

Effects of Mobility on Membership Estimation and Routing Services in Ad Hoc Networks

The use of MANETs (or Mobile Ad hoc NETworks) is becoming very popular. Power efficiency is a key issue in this type of network, as mobile devices usually rely on limited power supplies. One essential service, the routing protocol, employed to discover routes between nodes in the network, can greatly affect power consumption. Furthermore, many distributed applications require an additional membership service to keep track of the nodes that make up the system at any moment. In general, this information is not provided by routing services with the exception of the Optimised Link State Routing protocol (OLSR). The two services, routing and membership estimation, form a basic support to build other higher---level distributed services on ad hoc networks. To decrease the over-all power consumption these services should be optimized for the intended use of the network. In particular the degree of mobility can have an impact on the power consumption and performance of different approaches to routing and membership estimation. In this paper we present a study of two different approaches that combine a routing service with membership estimation. We compare the proactive OLSR with our own approach. Our approach consists of integrating a gossip-style failure detector with the reactive Dynamic Source Routing protocol (DSR). We present an analysis of the effects of mobility on the global performance and power consumption of the two approaches. We identify scenarios for which each approach is best suited.

Extended membership problem for open groups: specification and solution

A particular case of open group is that of a large–scale system where an unbounded and dynamically changing set of client nodes continuously connect and disconnect from a reduced and stable group of servers. We propose an extended specification of the membership problem, that takes into account the different consistency properties that both types of nodes require from the membership information in such a scenario. The specification is completely independent of any group communication layer, assuming only the presence of a local failure detector at each node. We also describe a membership service that satisfies the specification and sketch a proof for its properties.