Multi-Provider Virtual Network Embedding With Limited Information Disclosure

Abstract

Abstract-The ever-increasing need to diversify the Internet has recently revived the interest in network virtualization. Network virtualization carries a significant benefit to Service Providers, as it enables the deployment of network services within customized virtual networks (VN) that offer performance and reliability guarantees. VN embedding across multiple substrate providers creates the need for a layer of indirection which is fulfilled by VN Providers. VN Providers are expected to have very limited knowledge of the physical infrastructure, since substrate providers do not disclose their network topology and resource information. This entails significant implications on resource discovery and assignment.In this paper, we study the challenging problem of multidomain VN embedding with limited information disclosure (LID). In this context, we discuss the visibility of VN Providers on substrate network resources and question the suitability of topologies for VN request specifications. Our main contributions are as follows: (i) we present a traffic matrix based VN embedding framework that enables VN request partitioning under LID, and (ii) we conduct a feasibility study on VN embedding with LID compared to a "best-case" scenario where all information is available to VN Providers.

Multi-Provider Service Chain Embedding With Nestor

“…Graph partitioning. Authors in [29] and our previous work [21] address the problem of multi-provider VN embedding. Both papers propose a LP for graph partitioning across multiple infrastructure providers (InPs).…”

“…In contrast, NF-graph partitioning generates NFsubgraphs mappable to DCs, and thereby, requires DC-level topology abstractions. Furthermore, Nestor provides NF-graph partitioning optimizations both the client and the NFP (i.e., via the disclosure of link weights expressing NFP preferances for link and DC selection), as opposed to VN-graph partitioning in [29], [21], which is optimized only for the client (i.e., expenditure minimization).…”

“…Such embeddings should satisfy the objectives of clients (e.g., expenditure minimization) and providers (e.g., revenue maximization), whereas embedding methods should address the intricacies of multi-provider aspects (i.e., restrictions in the resource and network topology information disclosed by NFPs to third parties) [21]. Existing solutions for multi-provider virtual network (VN) mapping [21], [29] generate embeddings based on VN graphs with specific bandwidth demands on each edge. This highly abstract service model cannot represent service chain resource demands, given the fact that certain NFs (e.g., caches, redundancy elimination boxes) may compress or amplify traffic, raising the need for resource reservations substantially different from the demands specified in the chain.…”

Multi-Domain Virtual Network Embedding with Coordinated Link Mapping

“…The proxy VNP could get peering links location and resource information but intra-domain links cannot be assumed to be available to VNP [11].…”

“…This phenomenon will result in difficulty of later intra-domain mapping. In [11], a virtual node is first mapped to substrate peering node to determine the peering link and the InP it belongs to. This approach is suitable for traffic matrix based VN [16] but not topology based VN.…”