Network management currently undergoes changes towards more flexible network management. This trend is stimulated by Network Virtualization and Software Defined Networks (SDN) that emerged in recent years. These technologies allow networks to be run in a more flexible and cost efficient manner, e.g., by increasing network resource utilization and by decreasing operational costs. As an emerging topic, Network Function Virtualization (NFV) allows even further flexibility by migrating network functions (e.g., DHCP, PPPoE) from dedicated hardware to virtual machines running on commodity hardware. Virtualized network functions are appealing to network operators since they can be migrated and flexibly adapted to current demands.
The newly achieved flexibility in network management, particularly for NFV, opens a set of currently unresolved key questions concerning i) reliability, ii) service orchestration iii) function placement, and iv) performance. How to operate virtualized network functions in a reliable manner by providing redundancy and load balancing? Can virtualized network functions provide performance figures required for network operations and how can such virtualized services be benchmarked and compared? Where should network functions be placed to optimize the network subject to different design criteria? How can services be orchestrated? How can network monitoring be adapted to such flexible networks? This workshop aims at addressing these and similar questions in virtualized networks.
Topics of interest for submissions include, but are not limited to:
- SDN/NFV architectures, applications, and use cases
- Network monitoring and QoE
- Reliability of virtualized network functions
- SDN/NFV-based service orchestration
- SDN/NFV-based network deployment and management
- Business considerations and economic aspects
- SDN/NFV security
- Theoretical foundations of SDN/NFV networks
- Network Operating Systems and Languages
- SDN in Mobile and Wireless Networks
- Network Service Chaining
|Thursday, 12 March 2015
|SlidesWelcome message by the workshop chairs
|Keynote: Nicolai Leymann (Deutsche Telekom) (Chair: David Hausheer)
|SlidesChallenges and Opportunities: The Rise of Software Centric NetworkingWith SDN and NFV new networking paradigms are changing the way how networks are being build and operated. This presentation will give an overview about the challenges and opportunities service providers are facing during the adoption phase. It addresses uses cases, architecture blue prints and the migration from today’s networks towards an SDN/NFV driven network deployment.
Nicolai Leymann is for more than 15 years with Deutsche Telekom as a senior network architect for IP based networks. His major focus is the network and service architecture of the upcoming Next Generation Network of Deutsche Telekom based on a Broadband Network Gateway (“BNG”). He is responsible on the network architecture of the converged service edge into the Germany-based IP network of Deutsche Telekom. His work also includes the extension of classical network services into virtualized environments based on standard components, deploying SDN and NFV approaches.
He is active in the IETF and ONF and author of about 20 Internet-Drafts and RFCs.
|Session 1: NFV (Chair: Kalman Graffi)
|PDFSlidesOpen Elastic Provisioning of Hardware Acceleration in NFV EnvironmentsNetwork Functions Virtualization (NFV) is a new paradigm to move network tasks currently running on dedicated, vendor-specific hardware to elastic, virtualized environments, similar to IaaS cloud computing. A major challenge of NFV is to reach the performance known from dedicated hardware appliances, which often leverage ASIC, FPGA or NPU-based hardware acceleration to increase throughput and reduce delay. In this paper, a framework for elastic provisioning, programming and configuration of acceleration hardware for virtual network functions (VNFs) is proposed. Using this framework, VNFs can offload selected parts of their workload to heterogeneous acceleration processors, which may be shared among all VNF instances in the framework for improved resource utilization.
Leonhard Nobach (Technische Universität Darmstadt), David Hausheer (Technische Universität Darmstadt)
|PDFSlidesMD2-NFV: The Case for Multi-Domain Distributed Network Functions VirtualizationMotivated by the vision of Network Functions Virtualisation (NFV) spanning different administrative domains, this position paper makes the case for multi-domain, distributed NFV (MD2-NFV). To this end, we present MD2-NFV as a natural evolution of the NFV paradigm to deliver a distributed deployment of Virtualized Network Functions (VNFs) as a service. By means of three motivating use case scenarios, we discuss potential benefits and identify challenging features towards enabling advanced peering relationships between NFV domains.
Raphael Vicente Rosa (University of Campinas, Brazil), Mateus Augusto Silva Santos (University of Campinas, Brazil) and Christian Esteve Rothenberg (University of Campinas, Brazil)
|Session 2: SDN (Chair: Thomas Zinner)
|PDFSlidesLoad-Dependent Flow Splitting for Traffic Engineering in Resilient OpenFlow NetworksIn this paper we propose and investigate load- dependent load balancing for resilient OpenFlow networks. The objective is to spare extra capacity for the primary path of a traffic aggregate (flow) by accommodating excess traffic on its backup path. The contribution of the paper is manyfold. We explain existing OpenFlow features for traffic monitoring and dynamic flow splitting, combine them to implement load-dependent balancing of a flow’s traffic on its primary and backup paths, and propose three different load-dependent flow splitting (LDFS) policies. We develop a performance evaluation method to quantify the capacity for load-balancing and protection switching methods such that expected traffic can be accommodated for envisioned overload and failure scenarios. Finally, we assess the usefulness of LDFS by comparison with traffic-agnostic single- or multipath forwarding methods.
Wolfgang Braun (Universität Tübingen), Michael Menth (Universität Tübingen)
|PDFSlidesTowards a GPU SDN ControllerThe SDN concept of separating and centralizing the control plane from the data plane has provided more flexibility and programmability to the deployment of the networks. On the other hand, the separation of the planes has raised some scalability and performance questions, being that the SDN controller is the bottleneck. In this paper we present an implementation of a GPU SDN controller. The goal of this paper is to mitigate the scalability problem of the SDN controller by offloading all the packet inspection and creation to the GPU. Experimental evaluation shows that the controller is able to process 17 Million packets/s using off-the-shelf GPU’s.
Eduard Gibert Renart (Rutgers University), Eddy Z. Zhang (Rutgers University), Badri Nath (Rutgers University
|PDFSlidesCost Efficiency of SDN in LTE-based Mobile Networks: Case FinlandMobile data traffic is expected to increase 11-fold between 2013 and 2018. A more dynamic and flexible network is needed to cope with the growing data volume. Software defined networking (SDN) enabled by, for example, the OpenFlow protocol is a potential solution to improve the flexibility and agility of the mobile data network by decoupling the control plane from the user plane. By centralizing the control plane into data centers, the network elements may become more standardized, which lowers prices due to scale-benefits and competition. At the same time, SDN may ease network management and deployment, which further reduces mobile network operator’s (MNO’s) costs. On the other hand, data centers become more complex due to virtualization and increased control intelligence. To quantify the cost impact, the changes in capital (CAPEX) and operational expenditure (OPEX) from adding SDN into the LTE network are modeled in this paper by using a Finnish reference network. The quantitative results show that SDN reduces the network related annual CAPEX by 7.72% and OPEX by 0.31% compared to non-SDN LTE. These changes, though a small fraction of the total expenses of a Finnish MNO, may have a significant influence on the profit levels.
Nan Zhang (Aalto University), Heikki Hämmäinen (Aalto University)
|Session 3: Virtualization Aspects (Chair: Oliver Hohlfeld)
|PDFSlidesInvestigating Isolation between Virtual Networks in Case of Congestion for a Pronto 3290 SwitchPerformance isolation between virtual resources is one of the key features of network virtualization. It is typically realized by configuring queues with specific rate guarantees on the egress ports of network devices. The drawback of this architectural choice, however, is that traffic from several ingress ports may result in congestion on an egress port. Hence, the question arises to which extent isolation between virtual networks is realized in state-of-the-art hardware. This work aims at investigating whether congestion within one virtual network may affect the throughput performance of another virtual network. For that, measurements in a local testbed using a Pronto 3290 switch running an OpenFlow-enabling Pica8 firmware are performed.
Anh Nguyen-Ngoc (University of Würzburg), Stanislav Lange (University of Würzburg), Steffen Gebert (University of Würzburg), Thomas Zinner (University of Würzburg), Phuoc Tran-Gia (University of Würzburg), Michael Jarschel (Nokia München)
|PDFSlidesProtocol Virtualization Through Dynamic Network StacksNetwork operators move away from specialized solutions and towards generic hardware and virtualized environments, allowing them to scale faster and at lower cost. It also allows them to adapt and control their network via software such as OpenFlow.
Today, this control loop does not provide end-to-end adaptation and control. Mainly due to the fact that end devices, e.g., mobile phones, and the applications installed are not yet part of this control loop. To enable this, we must push virtualization to the end device.
In this paper, we introduce the Dynamic Network Stack (DyNS). DyNS is based on the idea of protocol virtualization. Hidden behind a virtual network interface, DyNS creates different network protocol stacks for each application. As requirements change, DyNS can seamlessly switch between network stacks. A first proof-of-concept implementation on Linux illustrates the feasibility by switching between UDP/IP and DCCP/IP with minimal overhead and transparency for the application.
Jens Heuschkel (TU Darmstadt - TK), Immanuel Schweizer (TU Darmstadt - TK), Torsten Zimmermann (RWTH Aachen - Comsys), Klaus Wehrle (RWTH Aachen - Comsys), Max Mühlhäuser (TU Darmstadt - TK)
|Workshop closing note
|Workshop Dinner (Brau & Bistro, Altmarkt, Cottbus [map])
Acceptance rate: We accepted 7 out of 18 submissions.
|October 20th, 2014 11:59 AM (extended)
|December 10th, 2014
|November 26th, 2014
|March, 12th, 2015
All submissions must be original, unpublished, and not considered elsewhere for publication. We invite submissions up to 5 pages long (10pt font, double column, IEEE format), including text, figures and references.
IEEE LaTeX and Microsoft Word templates, as well as formatting instructions, are available online on this website.
Each paper will undergo a thorough process of peer reviews by at least three members of the technical program committee. All papers accepted in the workshop will be included in the conference proceedings (submitted to IEEE Xplore®). Submission implies that at least one author will register and attend the workshop to present the publication if the paper is accepted.
- David Hausheer, TU Darmstadt, Germany
- Oliver Hohlfeld, RWTH Aachen University, Germany
- Thomas Zinner, Julius-Maximilian University Würzburg, Germany
- Marco Hoffmann (Nokia)
- Michael Jarschel (Nokia)
- Marcus Schöller (NEC, vice-chair IETF Reliability and Availability WG, ETSI ISG NFV)
- Fabian Schneider (NEC)
- Bernhard Ager (ETH Zürich)
- Gerhard Haßlinger (Deutsche Telekom)
- Phuoc Tran-Gia (Uni Würzburg)
- Stefan Schmid (TU Berlin / Telekom Innovation Laboratories)
- Marco Canini (Université catholique de Louvain)
- Laurent Vanbever (Princeton University)
- Thomas Magedanz (TU Berlin / Fraunhofer FOKUS)
- Steven Latre (Uni Antwerpen)
- Dariusz Bursztynowski (Orange Poland)
- Wolfgang Kellerer (TU München)
- Andreas Wundsam (Big Switch, ICSI / UC Berkeley)
- Michael Menth (Uni Tübingen)
- Dave Meyer (Brocade)
- Christian Esteve Rothenberg (CPqD - Campinas)
- Xenofontas Dimitropoulos (University of Crete & FORTH)
- Volker Hilt (Alcatel-Lucent Bell-Labs)