MULE: Multi-Layer Virtual Network Embedding Shihabur Rahman Chowdhury * , Sara Ayoubi * , Reaz Ahmed * , Nashid Shahriar * , Raouf Boutaba * , Jeebak Mitra , and Liu Liu * David R. Cheriton School of Computer Science, University of Waterloo {sr2chowdhury | sayoubi | r5ahmed | nshahria | rboutaba}@uwaterloo.ca Huawei Technologies Canada Research Center jeebak.mitra@huawei.com Huawei Technologies liuliu1@huawei.com Abstract—Network Virtualization (NV), considered as a key enabler for overcoming the ossification of the Internet allows multiple heterogeneous virtual networks to co-exist over the same substrate network. Resource allocation problems in NV have been extensively studied for single layer substrates such as IP or Optical networks. However, little effort has been put to address the same problem for multi-layer IP-over-Optical networks. The increasing popularity of multi-layer networks for deploying backbones combined with their unique characteristics (e.g., topological flexibility of the IP layer) calls for the need to carefully investigate the resource provisioning problems arising from their virtualization. In this paper, we address the problem of MU lti-L ayer virtual network E mbedding (MULE) on IP-over- Optical networks. We propose two solutions to MULE: an Integer Linear Program (ILP) formulation for the optimal solution and a heuristic to address the computational complexity of the optimal solution. We demonstrate through extensive simulations that on average our heuristic performs within 1.47× of optimal solution and incurs 66% less cost than the state-of-the-art heuristic. I. I NTRODUCTION Multi-layer IP-over-Optical networks are becoming a popu- lar choice among Infrastructure Providers (InPs) for deploying wide area networks [1]. Such multi-layer network typically consists of an optical substrate for the physical communication with an IP overlay on top [2]. This network model is being increasingly adopted for backbone networks as it offers the best of both worlds, i.e., the flexibility in addressing, resource allocation, and traffic engineering of IP networks along with the high capacity provided by optical networks. Despite their increasing popularity, research on addressing resource pro- visioning challenges for virtualizing such networks is still in its infancy. A classical resource provisioning problem in network virtualization is Virtual Network Embedding (VNE), which consists in establishing a Virtual Network (VN) on a Substrate Network (SN) with objectives such as minimizing resource provisioning cost [3], [4], maximizing the number of admitted VNs [5], etc. VNE has been extensively studied for single-layer SNs [6] with significantly lesser attention paid to the multi-layer network substrates [7]. The topological flexibility provided by multi-layer networks [8] poses some unique challenges for VNE and calls for new investigations. Similar to multi-layer networks, Mule, a hybrid species brings the best of two species together. Several deployment models exist for multi-layer IP-over- Optical networks [9] including but not limited to: (i) IP over Dense Wavelength Division Multiplexed (DWDM); (ii) IP over Optical Transport Network (OTN) over DWDM. DWDM networks have specific constraints such as wavelength conti- nuity for optical circuits and typically do not have transparent traffic grooming capabilities. A more favorable choice (also our choice of technology) is to deploy an OTN [10] over a DWDM network with advanced transport capabilities (e.g., traffic grooming without optical-electrical-optical conversion). The OTN in turn can be static, i.e., necessary interfaces on OTN nodes have been configured and the corresponding light paths in the DWDM layer have been lit to provision fixed bandwidth between OTN nodes. Or, the OTN can be dynamic, i.e., more bandwidth between OTN nodes can be provisioned by lighting new light paths in the DWDM. Clearly, the VNE problem for each of these scenarios requires dedicated explorations due to their unique constraints. As a first step towards addressing VNE for multi-layer networks, we limit the scope of this paper to the case of a static OTN and leave the other possible deployment scenarios for future investigation. Solving the VNE problem for multi-layer networks exhibits many unique challenges due to the topological flexibility offered by such networks. Concretely, although the OTN is fixed, the IP network is dynamic, i.e., new IP links can be established when needed by provisioning necessary capacity from the OTN. Such flexibility can be exploited if residual resources in the IP layer are insufficient to admit a new VN, or to reduce the cost of VN embedding by creating new IP links that reduce network diameter. Provisioning new IP links in optical networks has been a tedious and manual task with a long turnaround time. However, with the advances in optical networking technologies [11] and centralized optical control plane [12]–[15], such provisioning tasks are more and more automated. Even then, one should not abuse such capability to sporadically establish new IP links since it remains more expensive than embedding virtual links on existing IP links. In this regard, we are faced with the following challenges: (i) strike a balance between obtaining a low cost VN embedding while minimizing the establishment of new IP links; (ii) simultaneously decide on whether to create an IP link or not and its embedding in the OTN. 978-3-901882-98-2 c 2017 IFIP