A Priority-Based Multicast Flow Scheduling Method for a Collaborative Edge Storage Datacenter Network

Abstract

Edge storage, as a supplement to cloud storage, reduces latency by providing services in a timely and efficient manner near the source. In a collaborative edge storage datacenter network (CESN), not only does the edge storage datacenter (ESDC) that is closest to the user provide services, but multiple ESDCs work together to provide better services. In this collaborative work mechanism, different application session requests create large persistent multicast flows with diverse performance requirements. Existing multicast scheduling methods such as unicast shortest path (USP) and static single tree (SST) do not consider flow characteristics or performance requirements. In this paper, we first modeled the multicast flow scheduling problem in a CESN. The model is based on different types of flows with diverse network requirements. Then, we tailored a multicast flow scheduling method based on multiple-attribute decisionmaking and a genetic algorithm (MDGA). MDGA selects appropriate multicast routing paths for flows in a CESN by considering the requested flow types and network status. The experimental results show that the proposed MDGA method can balance network loads and reduce the average transmission delay for highpriority flows better than USP and SST.INDEX TERMS collaborative edge storage, datacenter network, multicast flow, multiple-attribute decision-making, genetic algorithm.

A Load-Aware Multistripe Concurrent Update Scheme in Erasure-Coded Storage System

“…where B p represents the set of residual bandwidths for each path P. It can be calculated using the SDN-based network measurement method in our previous work [27].…”

“…where D p represents the set of transmission delays for each path. It can be obtained by using the SDN-based network measurement method [27].…”

A New Method for Reconstructing Data Considering the Factor of Selected Provider Nodes Set in Distributed Storage System

“…Due to the characteristics of MDS code, the RS code storage system has the best storage efficiency. Compared with our previous work [38], we have research the write operation in distributed storage system in the multi-copy way, both the multi-copy and RS coding way should both conduct the multicast problem.…”

“…Ryu controller senses the topology information through the awareness module through the OpenFlow protocol, collects the link delay and bandwidth status using the delay and monitor modules, and packet_in the path between nodes in mininet through the forwarding module. Refer to our original working literature [38], and its system structure is shown in Figure 4 below.…”