Log Replication in Raft vs Kafka

Petrescu, Manuela-Andreea; Petrescu, Relly Victoria

doi:10.24193/subbi.2020.2.05

Cited by 4 publications

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The protocol used in case of network errors is that the leader should try indefinitely to send messages to the followers. Data consistency is guaranteed by timers usage, so the followers logs will contain only valid data [12].…”

Section: Leader Election Process In Zookeepermentioning

confidence: 99%

Leader Election in a Cluster using Zookeeper

Petrescu

2021

Studia UBB Informatica

Self Cite

View full text Add to dashboard Cite

This paper presents an algorithm for flexible and fast leader election in distributed systems using Apache Zookeeper for configuration management. The algorithm proposed in this paper is designed for applications that do not use symmetric nodes so they need a specialized election process or for applications that require a more flexible approach in the leader election process. The algorithm proposes a different approach as it allows assigning prioritizations for servers in the cluster that are candidates to become a leader. The algorithm is flexible as it takes into consideration during the leader election process of the different server settings and roles, network properties, communication latency or specific application requirements.

show abstract

Section: Leader Election Process In Zookeepermentioning

confidence: 99%

Leader Election in a Cluster using Zookeeper

Petrescu

2021

Studia UBB Informatica

Self Cite

View full text Add to dashboard Cite

show abstract

Replication in Raft vs Apache Zookeeper

Petrescu

2023

Advances in Intelligent Systems and Computing

View full text Add to dashboard Cite

Choosing Methodology of How to Manage Distributed Information Systems under Conditions of High Dynamics of Network Infrastructure

Klimenko

2022

Proceedings of the SWSU

View full text Add to dashboard Cite

Purpose of research. is to identify a distributed system control method that, based on known parameters, would reduce the resource consumption of computing devices. An information system resource is probability of failure-free operation (FFO), which decreases over time for each node, the faster the higher its load.Methods. Frequency of system reconfiguration becomes relatively high with high dynamics of edge layer of a network and the need for reconfiguration is unpredictable. Reducing the total time spent on reconfiguration allows to increase time spent on solving functional computing tasks of the system and thereby reducing nodes load. Reconfiguration time can be reduced both by reducing failure detection time in distributed system and by reducing searching time for new configuration. A method for reducing failure detection time is described. The applicability analysis of system management methods (centralized, distributed leader, decentralized) is carried out on the basis of obtained analytical evaluations of failure detection time by the system in control conditions through one or another method. A numerical experiment allows to distinguish areas of system parameters, where it is preferable to use the method with a distingods. Frequency of system reconfiguration becomes relatively high with high dynamics of edge layer of a network and the need for reconfiguration is unpredictable. Reducing the total time spent on reconfiguration allows to increase time spent on solving functional computing tasks of the system and thereby reducing nodes load. Reconfiguration time can be reduced both by reducing failure detection time in distributed system and by reducing searching time for new configuration. A method for reducing failure detection time is described. The applicability analysis of system management methods (centralized, distributed leader, decentralized) is carried out on the basis of obtained analytical evaluations of failure detection time by the system in control conditions through one or another method. A numerical experiment allows to distinguish areas of system parameters, where it is preferable to use the method with a distinuished leader.Results. The main result of this work is a method of choosing how to manage distributed information systems in conditions of high dynamics of network infrastructure, focused on reducing of resource consumption in computing devices.Conclusion. The system reconfiguration time could be reduced by selecting the most appropriate control method. In this case, the time for solving functional tasks of application is increased, the workload of computing nodes is reduced, and therefore FFO values are increased over the planning horizon.

show abstract

Log Replication in Raft vs Kafka

Cited by 4 publications

References 0 publications

Leader Election in a Cluster using Zookeeper

Leader Election in a Cluster using Zookeeper

Replication in Raft vs Apache Zookeeper

Choosing Methodology of How to Manage Distributed Information Systems under Conditions of High Dynamics of Network Infrastructure

Contact Info

Product

Resources

About