A model to select the right infrastructure abstraction for Service Function Chaining

Abstract

Abstract-Through network function virtualization (NFV), telecom providers aim to flexibly re-use generic-purpose hardware to provide services on-demand and in an agile way. Service function chaining is becoming the preferred model to describe the characteristics of the packet-processing network functions which, together, form these services. NFV allows for network function embedding freedom, creating new dynamics between providers and the users requesting services. Users want this freedom to optimise the performance of their requested services, while providers aim to optimise their resource cost with it. This trade-off is heavily influenced by how the available infrastructure is exposed to the users. In this paper, we present an infrastructure abstraction model for network, compute and storage resources that exposes the infrastructure in an abstracted manner. We use this abstraction to propose a solution for the placement freedom trade-off problem by studying its relation with metrics that capture both the user's and the provider's aspects. We conclude with a heuristic that determines the right abstraction for particular scenarios.

Resource Abstractions in NFV Management and Orchestration: Experimental Evaluation

“…Both orchestration layers tightly cooperate to realize the resource decisions when provisioning an end-to-end network service request, i.e, placing the VNFs and deploying the VLs. To this end, the lower-layer resource orchestrator(s) typically expose an abstracted view of their controlled domains to the higher-layer Service Orchestrator/User [1], [8]. This abstracted resource information has a domain specific scope, e.g., aggregated available PoPs' compute resources, virtualized network connectivity, etc.…”

“…In this model, two strategies can be adopted to abstract the LLs from a transport domain. In the first, the exposed LLs are explicitly induced from pre-calculated transport paths between pairs of domain edges [16] [10] [8]. Consequently, the resulting LLs inherit the physical path attributes, i.e., available path bandwidth, path delay, etc.…”

“…In addition to the cloud and network abstraction, the orchestration functions include the resource selection and allocation. These functions are essential to achieve a good trade-off between abstraction accuracy and efficient resource utilization [8]. To this end, in [10], we introduced two different operating modes in the Resource Layer orchestrator called Infrastructure Abstraction (InA) and Connectivity Service Abstraction (CSA).…”

Optimising microservice-based reliable NFV management & orchestration architectures

“…As (parts of) this state are being transmitted, one should make sure that the network is not overloaded. One way of dealing with this is using abstractions of the topology [9] which are considerably smaller.…”

Implementation of service function chaining control plane through OpenFlow

“…SFC Orchestrators to deploy SFCs as well as control their activity and make adjustments are introduced in [3]. The problem of allocating physical resources to data plane components of a SFC is addressed in [4], while a solution for the trade-off between optimized performances and resource cost in SFC deployments is presented in [5].…”