Reinhold Kröger

Recovery-management in the RelaX distributed transaction layer

Transactions are especially valuable in distributed systems, since they isolate the programmer from the effects of both concurrency and failures. In implementing transactions at the system level, flexibility has to be introduced into the transaction concept. Specifically, the premature release of data objects has to be addressed. To assure recoverability, resulting dependencies between transactions are stored by the system in a distributed data structure called a recovery graph. The storing of redundant information in the recovery graphs at the different sites reduces the complexity of the commit protocol, and a chase protocol used to abort transactions can be derived which excludes infinite chasing. The redundant information can be distributed almost for free because it can be piggybacked on messages. The new commit/abort protocols will be used in the RelaX project (reliable distributed applications support on Unix), which carries on work done at GMD in the PROFEMO project on distributed transaction mechanisms.<<ETX>>

Mutabor - An Intelligent Memory Management Unit for an Object Oriented Architecture Supporting Error Recovery

In the project PROFEMO (project for the design and implementation of a fault-tolerant multiprocessor system) objects are primarily regarded as an appropriate means to increase system reliability. Their contribution to reliability includes improved error detection, damage confinement, and recoverability features. This results from the well -defined object interface and the encapsulation aspect of objects. For error recovery, objects form meaningful entities for which recovery can be performed. This paper describes a memory management coprocessor which provides capability protected, typed objects by dedicated hard/firmware support, making the access to objects efficient and reliable. It also realizes the save/restore managent of recovery points for objects which forms the basic layer of the recovery mechanism.

Implementing fault-tolerance in a distributed system architecture

Atomic actions are widely accepted as a powerful general concept for providing consistency and error recovery. However, so far the necessary mechanisms to implement these properties severely confine interprocess communication and the degree of internal concurrency and thereby interfering with the main advantages of distributed systems. Based on the concept of an object oriented architecture an efficient error recovery strategy is presented which largely overcomes these constraints. To do so, we consider the premature release of objects, i.e. allowing access to them before the associated atomic action has committed. The atomic action concept is extended by separating the termination of such an action from the commitment procedure. This avoids possible delay of the affected participants. The redundancy, inherent in a distributed system, is exploited to run computations and necessary recovery activity concurrently.

Concept and Performance of the RelaX Reliable Broadcast Protocol

Reliable Broadcasting is a very important communication mechanism in fault- tolerant distributed systems. In the RelaX project, a reliable broadcast protocol has been developed and implemented, which uses only one acknowledgement in the fault-free case. It is resistent against site crashes and ensures that a broadcast message is transmitted to all operational sites. Moreover, it provides a global message ordering. An important point is that the protocol detects site faults, so that every site knows about all the other operational sites. The measured performance of the protocol shows its efficiency.

Implementierung eines RPC-Protokolls für eine objektorientierte Umgebung

In diesem Beitrag wird ein RPC-Protokoll beschrieben, das im Zusammenhang mit dem Projekt PROFEMO entworfen und implementiert wurde. Das Protokoll berucksichtigt die besonderen Bedurfnisse einer objektorientierten verteilten Umgebung. Die ersten Messungen zeigen, das das Protokoll diesen Bedurfnissen gerecht wird und mit ublichen Implementierungen in seiner Leistungsfahigkeit vergleichbar ist.