Felicián Németh

On QoS Support to Ofelia and OpenFlow

OpenFlow is the most promising architecture for future Software Defined Networks (SDNs). However, from the aspects of large-scale or carrier-grade networks, it still lacks some key components. For example, QoS (Quality of Service) provisioning is an indispensable part of such production networks. During the evolution of the OpenFlow standard, some QoS capabilities have been added to the protocol, however, even the latest version has only a limited and not well-defined QoS framework. Hence, integrated QoS support is missing in current OpenFlow experimental test beds including Ofelia. This paper describes a possible architectural extension to Ofelia in order to make it capable of running QoS related experiments. We summarize the initial tasks regarding the survey of QoS features and limitations of OpenFlow switches deployed in Ofelia islands and the performance tests needed to characterize these devices. For extending the feature palette of Ofelia, we propose a QoS management platform with full integration into the existing management framework. By means of this envisioned extension, QoS settings of the whole Ofelia test bed can be adjusted easily, in a user friendly fashion. Moreover, we walk through the main steps needed not only towards an integrated OpenFlow test bed with QoS support but towards a QoS architecture to OpenFlow.

Towards Unified Programmability of Cloud and Carrier Infrastructure

The rise of cloud services poses considerable challenges on the control of both cloud and carrier network infrastructures. While traditional telecom network services rely on rather static processes (often involving manual steps), the wide adoption of mobile devices including tablets, smartphones and wearables introduce previously unseen dynamics in the creation, scaling and withdrawal of new services. These phenomena require optimal flexibility in the characterization of services, as well as on the control and orchestration of both carrier and cloud infrastructure. This paper proposes a unified programmability framework addressing: the unification of network and cloud resources, the integrated control and management of cloud and network, the description for programming networked/cloud services, and the provisioning processes of these services. In addition proofs-of-concept are provided based on existing open source control software components.

Nonrate-proportional weighting of generalized processor sharing schedulers

This paper is concerned with GPS-based packet schedulers which are very important elements of guaranteed QoS networks. After recalling the basic properties of generalized processor sharing we introduce a novel approach to calculate tighter delay bounds than given in the former works of Parekh et al. (1993, 1994). This not only allows better utilization of networks resources but supports basis for arbitrary weighting of sessions. An efficient and numerically inexpensive algorithm for computing these bounds is also presented herein. Further we relax the rate-proportional weighting constraint and show that our delay bound calculation is applicable for any arbitrary (nonrate-proportional) weighted GPS system. Numerical examples demonstrate the flexibility of our algorithm in terms of adjusting delay and bandwidth easily between theoretical bounds. This approach is a very important step towards avoiding bandwidth-delay coupling which is a well-known problem of GPS-based rate-proportional servers.

ESCAPE: extensible service chain prototyping environment using mininet, click, NETCONF and POX

Mininet is a great prototyping tool which combines existing SDN-related software components (e.g., Open vSwitch, OpenFlow controllers, network namespaces, cgroups) into a framework, which can automatically set up and configure customized OpenFlow testbeds scaling up to hundreds of nodes. Standing on the shoulders of Mininet, we implement a similar prototyping system called ESCAPE, which can be used to develop and test various components of the service chaining architecture. Our framework incorporates Click for implementing Virtual Network Functions (VNF), NETCONF for managing Click-based VNFs and POX for taking care of traffic steering. We also add our extensible Orchestrator module, which can accommodate mapping algorithms from abstract service descriptions to deployed and running service chains.

OpenFlow Virtualization Framework with Advanced Capabilities

Due to its simplicity, transparency and performance, Flow Visor takes it all in todays virtualization tools for OpenFlow networks. We argue that this effectiveness comes at the price of intolerance towards diverse OpenFlow versions used simultaneously and also limited switch functionality. What is more Flow Visor based management frameworks cannot run and configure the network controllers. In this paper we present an integrated OpenFlow virtualization framework, that is capable of (i) running and managing multiple instances of OpenFlow switches with different forwarding capabilities and OpenFlow versions, (ii) running and configuring full controllers or network applications designed for controlling a virtual network under the management of the proposed framework, and (iii) configuring QoS in the network. Besides introducing the architecture we share implementation and deployment details in our prototype system.

SDN based testbeds for evaluating and promoting multipath TCP

Multipath TCP is an experimental transport protocol with remarkable recent past and non-negligible future potential. It has been standardized recently, however the evaluation studies focus only on a limited set of isolated use-cases and a comprehensive analysis or a feasible path of Internet-wide adoption is still missing. This is mostly because in the current networking practice it is unusual to configure multiple paths between the endpoints of a connection. Therefore, conducting and precisely controlling multipath experiments over the real “internet” is a challenging task for some experimenters and impossible for others. In this paper, we invoke SDN technology to make this control possible and exploit large-scale internet testbeds to conduct end-to-end MPTCP experiments. More specifically, we establish a special purpose control and measurement framework on top of two distinct internet testbeds. First, using the OpenFlow support of GÉANT, we build a testbed enabling measurements with real traffic. Second, we design and establish a publicly available large-scale multipath capable measurement framework on top of PlanetLab Europe and show the challenges of such a system. Furthermore, we present measurements results with MPTCP in both testbeds to get insight into its behavior in such not well explored environment.

OSPF for Implementing Self-adaptive Routing in Autonomic Networks: A Case Study

Autonomicity, realized through control-loop structures operating within network devices and the network as a whole, is an enabler for advanced and enriched self-manageability of network devices and networks. In this paper, we argue that the degree of self-management and self-adaptation embedded by design into existing protocols needs to be well understood before one can enhance or integrate such protocols into self-managing network architectures that exhibit more advanced autonomic behaviors. We justify this claim through an illustrative case study: we show that the well-known and extensively used intra-domain IP routing protocol, OSPF, is itself a quite capable self-managing entity, complete with all the basic components of an autonomic networking element like embedded control-loops, decision-making modules, distributed knowledge repositories, etc. We describe these components in detail, concentrating on the numerous control-loops inherent to OSPF, and discuss how some of the control-loops can be enriched with external decision making logics to implement a truly self-adapting routing functionality.

Towards SmartFlow: case studies on enhanced programmable forwarding in OpenFlow switches

The limited capabilities of the switches renders the implementation of unorthodox routing and forwarding mechanisms as a hard task in OpenFlow. Our high level goal is therefore to inspect the possibilities of slightly smartening up the OpenFlow switches. As a first step in this direction we demonstrate (with Bloom filters, greedy routing and network coding) that a very limited computational capability enables us to natively support experimental technologies while preserving performance. We distribute the demos in source files and as a ready-to-experiment VM image to promote further improvements and evaluations.

One tool to rule them all: a modular troubleshooting framework for SDN (and other) networks

Due to the heterogeneous and distributed nature of computer networks, the detection of misconfigurations and software/hardware failures is frequently reported to be notoriously non-trivial. The advent of SDN complicates the situation even more, since besides troubleshooting, the problem of finding software bugs in controller/switch/VNF implementations also has to be solved. Today a wealth of general and SDN-specific troubleshooting tools are available which are usually tailored to identify network-related errors and bugs of a particular nature. In this paper we define a troubleshooting framework which can assemble many of these tools in a single platform and makes possible to flexibly combine them. As we see network operators and SDN developers execute similar tasks anyway, e.g. combine ping, traceroute and tcpdump (or more complex tools) manually to see what is going on in the network. Our framework can ease their work by consolidating the available troubleshooting tools in a flexible and automated manner.

Call admission control in generalized processor sharing schedulers with tight deterministic delay bounds

Generalized processor sharing (GPS) is reported as the most important ideal fluid scheduling discipline, which many practical packet scheduling algorithms (WFQ, WF^2Q) are built on. However, all of them bear the bandwidth-delay coupling property due to the commonly used rate-proportional weighting of sessions and offer too loose deterministic delay bounds. These deficiencies lead to poor network utilization and preclude the development of efficient call admission control (CAC) schemes. In this paper we address these problems by proposing an algorithm for tight individual session delay bound computation and several CAC algorithms for session resource assignment in a bandwidth-delay decoupled GPS system. Besides the analytical framework numerical examples as well as the complexity analysis of different CAC algorithms are also presented.