Stanislav Lange

Heuristic Approaches to the Controller Placement Problem in Large Scale SDN Networks

Software Defined Networking (SDN) marks a paradigm shift towards an externalized and logically centralized network control plane. A particularly important task in SDN architectures is that of controller placement, i.e., the positioning of a limited number of resources within a network to meet various requirements. These requirements range from latency constraints to failure tolerance and load balancing. In most scenarios, at least some of these objectives are competing, thus no single best placement is available and decision makers need to find a balanced trade-off. This work presents POCO, a framework for Pareto-based Optimal COntroller placement that provides operators with Pareto optimal placements with respect to different performance metrics. In its default configuration, POCO performs an exhaustive evaluation of all possible placements. While this is practically feasible for small and medium sized networks, realistic time and resource constraints call for an alternative in the context of large scale networks or dynamic networks whose properties change over time. For these scenarios, the POCO toolset is extended by a heuristic approach that is less accurate, but yields faster computation times. An evaluation of this heuristic is performed on a collection of real world network topologies from the Internet Topology Zoo. Utilizing a measure for quantifying the error introduced by the heuristic approach allows an analysis of the resulting trade-off between time and accuracy. Additionally, the proposed methods can be extended to solve similar virtual functions placement problems which appear in the context of Network Functions Virtualization (NFV).

Specialized Heuristics for the Controller Placement Problem in Large Scale SDN Networks

The Software Defined Networking (SDN) concept introduces a paradigm shift in the networking world towards an externalized control plane which is logically centralized. When designing an SDN-based WAN architecture, it is of vital importance to find a feasible solution to the controller placement problem, i.e., to decide where to position a limited amount of resources within the network. In addition to time-independent constraints regarding aspects like scalability, resilience, and control plane communication delays, dynamically changing network conditions like traffic patterns or bandwidth demands need to be considered as well. Consequently, such dynamic environments call for a regular and fast recalculation of placements in order to adapt to the current situation in a timely manner. While an exhaustive evaluation of all possible solutions can be performed within a practically feasible time frame for small and medium-sized networks, such an approach is out of scope for large problem instances which have significantly higher time and memory requirements. Therefore, this work investigates a specialized heuristic, which takes into account a particular set of optimization objectives and returns solutions representing the possible trade-offs between them. Due to its low computation time and acceptable margin of error, this heuristic can be employed by automatic decision systems operating in dynamic environments.

Performance benchmarking of a software-based LTE SGW

Network Functions Virtualization (NFV) is a concept that aims at providing network operators with benefits in terms of cost, flexibility, and vendor independence by utilizing virtualization techniques to run network functions as software on commercial off-the-shelf (COTS) hardware. In contrast, prior solutions rely on specialized hardware for each function. Performance evaluation of such systems usually requires a dedicated testbed for each individual component. Rather than analyzing these proprietary black-box components, Virtualized Network Functions (VNFs) are pieces of software that run on COTS hardware and whose properties can be investigated in a generic testbed. However, depending on the underlying hardware, operating system, and implementation, VNFs might behave differently. Therefore, mechanisms for the performance evaluation of VNFs should be similar to benchmarking of software, where different implementations are compared by applying them to predefined test cases and scenarios. This work presents a first step towards a benchmarking framework for VNFs. Given two different implementations of a VNF that acts as LTE Serving Gateway (SGW), influence factors and key performance indicators are identified and a comparison between the two mechanisms is drawn.

Investigating the impact of network topology on the processing times of SDN controllers

Software Defined Networking (SDN) introduces the concept of logically-centralized controllers in charge of managing the forwarding behavior of network elements. The new possibilities enabled through the centralization of control logic come with a certain risk: The controller might become a performance bottleneck. Therefore, ensuring sufficient controller performance is one of the crucial tasks prior to a successful SDN deployment. Furthermore, fine-grained traffic engineering, e.g., to achieve higher link utilization, results in a higher frequency of requests that are sent to the controller, which leads to an increased controller load. It is therefore important to analyze the capabilities of SDN controllers prior to deployment. This paper investigates two software implementations, the OpenDaylight and Ryu controllers. The control message throughput of different controllers has been studied several times already; however, it is not yet known what influence the number and topology of connected switches have. This paper investigates this influence in detail for a fat-tree data center topology and a WAN topology as well as 261 topologies with varying characteristics from the Internet Topology Zoo.

Performance evaluation mechanisms for FlowMod message processing in OpenFlow switches

Network operators can benefit in terms of flexibility, cost, and vendor-independence when adopting the Software Defined Networking (SDN) paradigm. In many scenarios, the SDN controller orders the installation of new flow table entries in the switches it manages. Since such operations are handled in the slow path of the switches, the corresponding processing times constitute an important performance indicator for switches. This work focuses on a comparison of two mechanisms for evaluating the performance of OpenFlow switches with respect to the processing time of FlowMod messages. These mechanisms are characterized by different degrees of accuracy, cost, complexity, and the capability of performing measurements at run time. The first mechanism is based on the Spirent C1 dedicated testing platform, while the other uses a software module for the OpenDaylight controller. We assess their capabilities with respect to the abovementioned characteristics and quantify their accuracy by means of wiretaps that provide a ground truth regarding the measured processing times. By using three different switches in the experiments, it is possible to distinguish between hardware specific side-effects and general phenomena.

Evaluation of Video Quality Monitoring Based on Pre-computed Frame Distortions

In the past decade, video streaming has taken over a large part of the current Internet traffic and more and more TV broadcasters and network providers extend their portfolio of video streaming services. With the growing expectations of video consumers with respect to the service quality, monitoring is an important aspect for network providers to detect possible performance problems or high network load. In parallel, emerging technologies like software defined networking or network virtualization introduce support for specialized networks which allow enhanced functionality in the network. This development enables more sophisticated monitoring techniques in the specialized networks which use knowledge about the video content to better predict the service quality at consumers. In this work, we present a SSIM-based monitoring technique and compare it with the current state-of-the-art which infers the service quality from the monitored packet loss. We further show how network conditions like packet loss or bursts influence the two different monitoring techniques.

Performance Modeling of Softwarized Network Functions Using Discrete-Time Analysis

The softwarization of networks promises cost savings and better scalability of network functions by moving functionality from specialized devices into commercial off-the-shelf hardware. Generalized computing hardware offers many degrees for adjustment and tuning, which can affect performance and resource utilization. One of these adjustments are the interrupt moderation techniques implemented by modern network interface cards and operating systems. Using these, an administrator can optimize either lower latencies or lower CPU overhead for processing of network traffic. In this work, an analytical model that allows computing relevant performance metrics like the packet processing time and the packet loss for generic virtualized network functions running on commodity hardware is developed. The applicability of the model is shown by comparing its outcome with measurements conducted in a local testbed featuring a VNF that acts as an LTE Serving Gateway (SGW). Based on this model, impact factors like the average packet interarrival time, the interarrival time distribution, and the duration of the interrupt aggregation interval are studied.

Automated Decision Making Methods for the Multi-objective Optimization Task of Cloud Service Placement

The network functions virtualization (NFV) paradigm provides advantages with respect to aspects like flexibility, costs, and scalability of networks. However, management and orchestration of the resulting networks also introduce new challenges. The placement of services and virtualized network functions (VNFs) is a multi-objective optimization task that confronts operators with a multitude of possible solutions that are incomparable among each other. The goal of this work is to investigate mechanisms that enable automated decision making between such multi dimensional solutions. To this end, we investigate techniques from the domain of multi attribute decision making that aggregate the performance of placements to a single numeric score. A comparison between resulting rankings of placements shows that many techniques produce similar results. Hence, placements that achieve good rankings according to many approaches might be viable candidates in the context of automated decision making.

Investigating isolation between virtual networks in case of congestion for a Pronto 3290 switch

Performance 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.

A Multi-objective Heuristic for the Optimization of Virtual Network Function Chain Placement

The Network Functions Virtualization (NFV) paradigm offers network operators benefits in terms of cost efficiency, vendor independence, as well as flexibility and scalability. However, in order to profit most from these features, new challenges in the area of management and orchestration of the virtual network functions (VNFs) need to be addressed.,,In particular, this work deals with the VNF chain placement problem (VNFCP). For a given network situation, the task consists of determining the number, location, and assignment of VNF instances and the routing of demands. At the same time, several metrics like CPU utilization and the delay of individual flows need to be taken into account. For applicability in networks with dynamically changing conditions, algorithms need to explore the solution space of this NP-hard problem in a timely manner.,,The contribution of this work is threefold: firstly, we design MO-VNFCP, a multi-objective heuristic for the VNFCP. Secondly, we investigate the convergence behavior of the algorithm in a case study. Finally, we provide a comparison between the proposed algorithm and an alternative approach from literature.