Distributed System Level Diagnosis Research Articles

A generalized model for distributed comparison-based system-level diagnosis

This work introduces a new system-level diagnosis model and an algorithm based on this model: Hi-Comp (Hierarchical Comparison-based Adaptive Distributed System-Level Diagnosis algorithm). This algorithm allows the diagnosis of systems that can be represented by a complete graph. Hi-Comp is the first diagnosis algorithm that is, at the same time, hierarchical, distributed and comparison-based. The algorithm is not limited to crash fault diagnosis, because its tests are based on comparisons. To perform a test, a processor sends a task to two processors of the system that, after executing the task, send their outputs back to the tester. The tester compares the two outputs; if the comparison produces a match, the tester considers the tested processors fault-free; on the other hand, if the comparison produces a mismatch, the tester considers that at least one of the two tested processors is faulty, but can not determine which one. Considering a system of N nodes, it is proved that the algorithm's diagnosability is (N-1) and the latency is log2N testing rounds. Furthermore, a formal proof of the maximum number of tests required per testing round is presented, which can be O(N³). Simulation results are also presented.

A generalized model for distributed comparison-based system-level diagnosis

Abstract This work introduces a new system-level diagnosis model and an algorithm based on this model: Hi-Comp (Hierarchical Comparison-based Adaptive Distributed System-Level Diagnosis algorithm). This algorithm allows the diagnosis of systems that can be represented by a complete graph. Hi-Comp is the first diagnosis algorithm that is, at the same time, hierarchical, distributed and comparison-based. The algorithm is not limited to crash fault diagnosis, because its tests are based on comparisons. To perform a test, a processor sends a task to two processors of the system that, after executing the task, send their outputs back to the tester. The tester compares the two outputs; if the comparison produces a match, the tester considers the tested processors fault-free; on the other hand, if the comparison produces a mismatch, the tester considers that at least one of the two tested processors is faulty, but can not determine which one. Considering a system of N nodes, it is proved that the algorithm’s diagnosability is (N-1) and the latency is log2N testing rounds. Furthermore, a formal proof of the maximum number of tests required per testing round is presented, which can be O(N3). Simulation results are also presented.

An Isochronous Testing Strategy for Hierarchical Adaptive Distributed System-Level Diagnosis

Distributed System-level diagnosis allows the fault-free components of a fault-tolerant distributed system to determine which components of the system are faulty and which are fault-free. The time it takes for nodes running the algorithm to diagnose a new event is called the algorithm's latency. In this paper we present a new distributed system-level diagnosis algorithm which presents a latency of O(log N) testing rounds, for a system of N nodes. A previous hierarchical distributed system-level diagnosis algorithm, Hi-ADSD, presents a latency of O(log2 N) testing rounds. Nodes are grouped in progressively larger logical clusters for the purpose of testing. The algorithm employs an isochronous testing strategy that forces all fault-free nodes to execute tests on clusters of the same size each testing round. This strategy is based on two main principles: a tested node must test its tester in the same rounds a node only accepts tests according to a lexical priority order. We present formal proofs that the algorithm's latency is at most 2log N – 1 testing rounds and that the testing strategy of the algorithm leads to the execution of isochronous tests. Simulation results are shown for systems of up to 64 nodes.

A hierarchical adaptive distributed system-level diagnosis algorithm

Consider a system composed of N nodes that can be faulty or fault-free. The purpose of distributed system-level diagnosis is to have each fault-free node determine the state of all nodes of the system. This paper presents a Hierarchical Adaptive Distributed System-level Diagnosis (Hi-ADSD) algorithm, which is a fully distributed algorithm that allows every fault-free node to achieve diagnosis in, at most, (log/sub 2/ N)/sup 2/ testing rounds. Nodes are mapped into progressively larger logical clusters, so that tests are run in a hierarchical fashion. Each node executes its tests independently of the other nodes, i.e., tests are run asynchronously. All the information that nodes exchange is diagnostic information. The algorithm assumes no link faults, a fully-connected network and imposes no bounds on the number of faults. Both the worst-case diagnosis latency and correctness of the algorithm are formally proved. As an example application, the algorithm was implemented on a 37-node Ethernet LAN, integrated to a network management system based on SNMP (Simple Network Management Protocol). Experimental results of fault and repair diagnosis are presented. This implementation by itself is also a significant contribution, for, although fault management is a key functional area of network management systems, currently deployed applications often implement only rudimentary diagnosis mechanisms. Furthermore, experimental results are given through simulation of the algorithm for large systems of 64 nodes and 512 nodes.

Information dissemination in distributed systems with faulty units

Consider a network consisting of units connected by links in which some units could be faulty. Suppose each unit has a message which must be transmitted to all other (fault-free) units. We present an algorithm for doing this in a network operating in a fully distributed manner that requires at most 3n logn+O(n) message transmissions by fault-free units. Among other things, our result can be used to devise an algorithm for distributed system level diagnosis which is more efficient than the best currently known algorithm for this purpose.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>