Fast and efficient network service embedding method with adaptive offloading to the edge

Abstract

Abstract-Next generation networks and applications have recently drawn the attention of many researchers in both academia and industry. The plethora of service orchestration, embedding and scheduling solutions being developed indicates that efficient resource utilization in the cloud and edge/fog architecture is of crucial importance in order to exploit the great economic potential of this infrastructure. In this paper we propose a novel service orchestration approach that aims for speed and quality in terms of service provisioning: our embedding algorithm instantly deploys end-to-end delay-constrained service graphs while regularly offloads the most burdened parts of the infrastructure applying a cost-aware VNF migration strategy. In this sense we propose a hybrid orchestration approach, which unites the advantages of online heuristic and offline optimizing service orchestration methods, with the goal of obtaining a system design that provides fast service placement decisions and efficient long-term operation. By an exhaustive evaluation of our orchestration system on core, cloud and edge network topologies, we show its benefits in efficiency and the collateral cost of migrations.

Advanced Computation Capacity Modeling for Delay-Constrained Placement of IoT Services

“…The algorithm framework has been applied to data-plane component orchestration, where the input structures are translated to the general service graph structure, and the heuristic is tuned to the domain-specific needs [27]. Addressing the network operation scenarios, where service requests continuously arrive beyond 0-day deployments, we have extended our framework with hybrid online-offline optimization capability [28]. Most recently, the algorithm has been integrated with our multi-operator 5G proof-of-concept architecture [19], and its advanced features, such as path anti-affinity, has been studied by complexity theory methodologies.…”

“…In our earlier work we used this greedy heuristic service placement algorithm framework in various different settings, including general service topology structure [26], data plane resource orchestration [27], hybrid online-offline operational setting [28], and integrated it into a 5G network architecture supporting advanced resource (anti)-affinity constraints [19]. Our greedy heuristic framework has served as basis for several patents submitted by Ericsson, patenting VNF decomposition option selection, elastic substrate capacity management and VNF interconnection constraints (Patents pending).…”

“…Link throughput capacities and requirements are not modeled in this paper as we focus on node resource abstraction. Such placement problem variants can be found in our previous work [ 19 , 26 , 27 , 28 ] and the literature [ 20 , 21 ].…”

“…We note that our results on the substrate graph model are general and independent of the service graph structure. Previously published and applied versions of the greedy backtrack approach are capable of handling general service structures [ 19 , 26 , 27 , 28 ]. The function generates hosting candidates and filters them according to the constraints, as shown in Formulation 1.…”

Descarga adaptativa em ambiente com névoa heterogênea: estudo de caso para a área da saúde

“…Apesar dos benefícios da computação em névoa, deve-se levar em consideração que este tipo de infraestrutura possui recursos limitados. Németh et al (2018) propuseram um método centralizado para inclusão de serviços que combina um programa linear inteiro com um algoritmo de mapeamento guloso. O mecanismo de orquestração de recursos proposto usa recursos computacionais e de rede em ambientes de computação de névoa emergentes.…”