Abstract
Decentralized clustering of modern information technology is widely adopted in various fields these years. One of the main reason is the features of high availability and the failure-tolerance which can prevent the entire system form broking down by a failure of a single point. Recently, toolkits such as Akka are used by the public commonly to easily build such kind of cluster. However, clusters of such kind that use Gossip as their membership managing protocol and use link failure detecting mechanism to detect link failures cannot deal with the scenario that a node stochastically drops packets and corrupts the member status of the cluster. In this paper, we formulate the problem to be evaluating the link quality and finding a max clique (NP-Complete) in the connectivity graph. We then proposed an algorithm that consists of two models driven by data from application layer to respectively solving these two problems. Through simulations with statistical data and a real-world product, we demonstrate that our algorithm has a good performance.
Highlights
Clustering technologies leverage a set of connected computers to work as a single system [1], which improves performance, fault-tolerance and scalability of the system
As the φ FD and other failure detection mechanisms proposed in [5,6,7,8,9,10,11,12] are not suitable for the scenario of stochastic packet loss, we propose an algorithm that can estimate the severity of packet loss of a link between two nodes based on the statistical information of TCP protocol, the round-trip-time (RTT)
OpenDaylight supports the feature of constructing a SDN controller cluster to provide most of the advantages brought by distributed systems
Summary
Clustering technologies leverage a set of connected computers to work as a single system [1], which improves performance, fault-tolerance and scalability of the system. It is extremely important in areas such as sensor networking, clouding computing, centralized controlling, etc. Compared with centralized clustering technology, the decentralized cluster has many advantages such as no single point bottleneck, no single point of failure, more flexibility [2]. It faces many challenges, one of which is failure detection [3]. In decentralized clusters there is no fixed supervisor who is responsible for failure detection and troubleshooting, leading to a more complicated failure detecting problem
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
