Autonomous Multi-dimensional Slicing for Large-Scale Distributed Systems

Abstract

Abstract. Slicing is a distributed systems primitive that allows to autonomously partition a large set of nodes based on node-local attributes. Slicing is decisive for automatically provisioning system resources for different services, based on their requirements or importance. One of the main limitations of existing slicing protocols is that only single dimension attributes are considered for partitioning. In practical settings, it is often necessary to consider best compromises for an ensemble of metrics.In this paper we propose an extension of the slicing primitive that allows multi-attribute distributed systems slicing . Our protocol employs a gossip-based approach that does not require centralized knowledge and allows self-organization. It leverages the notion of domination between nodes, forming a partial order between multi-dimensional points, in a similar way to SkyLine queries for databases. We evaluate and demonstrate the interest of our approach using large-scale simulations.

Fluidify: Decentralized Overlay Deployment in a Multi-cloud World

“…To mitigate such situations, we use simulated annealing (taking inspiration from recent works on epidemic slicing [19]), resulting in a decentralized protocol for the deployment of overlay networks that is generic, efficient and scalable. 3 The Fluidify algorithm…”

“…To mitigate the risk of local minimums, we extend it with simulated annealing [19], which allows two nodes to be swapped even if there is an increase in the cost function. We call the resulting protocol Fluidify (SA), shown in Figure 5.…”

“…6) where each node considers a set of random nodes from N for a possible swap. The other is inspired from epidemic slicing [9,19], and only considers the physical neighbors of a node n for a possible swap (Slicing (SA), in Figure. 8).…”

“…Here nodes are given id values and numerical input values. Nodes exchange these input values so that in the end the input of rank k is located at the node with id k. Slicing method [19] [9] was made use of in resource allocation. Specific attributes of network(memory, bandwidth, computation power) are taken into account to partition the network into slices.…”

“…It has higher scalability and robustness and less maintenance cost compared to other approaches. The simulated annealing and slicing approach is motivated mainly by the works [19], [6], [9]. But these works concentrated mainly on improving a single network service while we concentrate on a generalized solution that can significantly improve all the network services.…”