Networking and SDN

We deal in projects with the topic of Access Networks and Software-Defined Networking (SDN). In particular we consider availability, reliability and serviceability for data center and WAN networks.

Journal Article: Analytical Model for SDN Signaling Traffic and Flow Table Occupancy and its Application for Various Types of Traffic

An extended version of our paper "Analytic Model for SDN Controller Traffic and Switch Table Occupancy" will be published in the next issue of the IEEE Transactions on Network and Service Management.

Mehr

Best Paper Award auf der CNSM 2016

At ACM/IEEE ComSoc Conference CNSM 2016, the team by Christopher Metter, Michael Seufert, Florian Wamser, Thomas Zinner and Phuoc Tran-Gia received the Best Paper Award.

Mehr

Bachelor Thesis: "Assessment of the ONOS SDN Controller in Case of Network Errors"

In this Bachelor Thesis we analyse the reconvergence performance of the ONOS controller for multiple network errors, e.g. packet loss on a data plane link.

Mehr

Paper: Analytic Model for SDN Controller Traffic and Switch Table Occupancy

Our Paper "Analytic Model for SDN Controller Traffic and Switch Table Occupancy" will be presented at CNSM 2016 in Montreal

Mehr

Related Projects

SDN-RAS² (Fast Failure Detection in Software-Defined Networks)

In systems design, a fail-fast system is one which immediately reports at its interface any condition that is likely to indicate a failure. Such designs often check the system's state at several points in an operation, so any failures can be detected early.

The objective of the project is the study and analysis of fast fault detection and localization techniques for SDN.

SDN-RAS (SDN - Reliabilty, Availabilty & Serviceability Challenges)

The project focuses on the identification and analysis of the top reliability, availability and serviceability challenges of current SDN mechanisms, protocols, and architectures.

Related Research Areas

Research Area: Access Networks

Study on Wifi Offloading: WiFi Offloading is a current trend to cope with the demands of mobile users and the load on cellular networks. It allows providers to handle the traffic in well-dimensioned fixed networks and thereby save costs. In addition, end users can benefit from higher throughput and avoid exceeding their data plan.

Application Aware Networking: Application-aware resource management is the approach to tailor access networks to have characteristics beneficial for the running applications and services. This is achieved through the monitoring and integration of key performance indicators from the application layer within the network resource management.

Research Area: Network Functions Virtualization - Research on Performance Evaluation

Network Function Virtualization (NFV) has become an emerge network technology in recent years. It has received a noticeable interest of not only researchers, network providers but also a huge market. NFV aims to reduce significantly the Capital Expenses (CAPEX) and Operating Expenses (OPEX) by leveraging virtualization technology to consolidate network functions running on equipment onto high volume commodity servers.

Considering this field of study, we are interested in the NFV architecture and its components. For example, the NFV infrastructure requires optimized methods to dynamically allocate virtual resources for VNFs. On the other hand, NFV management and orchestration needs a VNF manager/placement engine to instantiate the VNFs in network with regard to energy saving, less traffic utilization, etc. On the top of the architecture, the VNFs must perform at least as efficient as the legacy functions. We are focusing on evaluating the VNF performance in comparison with a legacy function. Other topics may relate to the trade-off between hardware utilization and the performance of VNFs and traffic modeling for virtual network functions chaining.

Publications

Projects SDN-RAS & SDN-RAS²

Towards an Active Probing Extension for the ONOS SDN Controller. Metter, Christopher; Burger, Valentin; Hu, Zheng; Pei, Ke; Wamser, Florian. In 2018 28th International Telecommunication Networks and Applications Conference (ITNAC) (ITNAC 2018), p. 8. Sydney, Australia, 2018.
- [ Abstract ]
- [ BibTeX ]
- [ EndNote ]
Network monitoring is a complex task and is always a trade-off between granularity of information and the performance impact of the monitoring itself on the network. SDN controllers, such as the ONOS SDN controller make this challenge easier as they can supply centralized information over the whole network. In our previous work we mathematically analyzed the built-in detection mechanism of ONOS and revealed a lack of detection performance and the vulnerability of this process to jitter. These results were also verified by measurements. In this paper we present an Active Probing extension for the ONOS SDN controller that overcomes these shortcomings by emitting probing packets that are transmitted through the network. Statistics over a history of those packets allow for the detection of packet loss. The evaluation by the means of measurements proves the benefits in terms of detection performance and controller load of this application. Furthermore, our extension is able to detect jitter being present in the network and to automatically adapt the probing process to these conditions.

@inproceedings{info3-inproceedings-2018-34, abstract = {Network monitoring is a complex task and is always a trade-off between granularity of information and the performance impact of the monitoring itself on the network. SDN controllers, such as the ONOS SDN controller make this challenge easier as they can supply centralized information over the whole network. In our previous work we mathematically analyzed the built-in detection mechanism of ONOS and revealed a lack of detection performance and the vulnerability of this process to jitter. These results were also verified by measurements. In this paper we present an Active Probing extension for the ONOS SDN controller that overcomes these shortcomings by emitting probing packets that are transmitted through the network. Statistics over a history of those packets allow for the detection of packet loss. The evaluation by the means of measurements proves the benefits in terms of detection performance and controller load of this application. Furthermore, our extension is able to detect jitter being present in the network and to automatically adapt the probing process to these conditions.}, address = {Sydney, Australia}, author = {Metter, Christopher and Burger, Valentin and Hu, Zheng and Pei, Ke and Wamser, Florian}, booktitle = {2018 28th International Telecommunication Networks and Applications Conference (ITNAC) (ITNAC 2018)}, keywords = {can}, month = 11, pages = 8, title = {Towards an Active Probing Extension for the {ONOS} {SDN} Controller}, year = 2018 }

%0 Conference Paper %1 info3-inproceedings-2018-34 %A Metter, Christopher %A Burger, Valentin %A Hu, Zheng %A Pei, Ke %A Wamser, Florian %B 2018 28th International Telecommunication Networks and Applications Conference (ITNAC) (ITNAC 2018) %C Sydney, Australia %D 2018 %P 8 %T Towards an Active Probing Extension for the ONOS SDN Controller %X Network monitoring is a complex task and is always a trade-off between granularity of information and the performance impact of the monitoring itself on the network. SDN controllers, such as the ONOS SDN controller make this challenge easier as they can supply centralized information over the whole network. In our previous work we mathematically analyzed the built-in detection mechanism of ONOS and revealed a lack of detection performance and the vulnerability of this process to jitter. These results were also verified by measurements. In this paper we present an Active Probing extension for the ONOS SDN controller that overcomes these shortcomings by emitting probing packets that are transmitted through the network. Statistics over a history of those packets allow for the detection of packet loss. The evaluation by the means of measurements proves the benefits in terms of detection performance and controller load of this application. Furthermore, our extension is able to detect jitter being present in the network and to automatically adapt the probing process to these conditions.
Evaluation of the Detection Capabilities of the ONOS SDN Controller. Metter, Christopher; Burger, Valentin; Hu, Zheng; Pei, Ke; Wamser, Florian. In 7th IEEE International Conference on Communications and Electronics (IEEE ICCE 2018), pp. 96–101. Hue, Vietnam, 2018.
- [ Abstract ]
- [ BibTeX ]
- [ EndNote ]
The process of monitoring the network, especially for larger ones, is very complex and contains many pitfalls. For instance the balance between granularity of information and their performance impact on the network. With the help of SDN this challenge can become easier, as it offers new methods, mechanisms and opportunities. One of the current most important Open Source controllers is the ONOS SDN controller. According to its developers it is a production ready controller that offers high availability due to its logically centralized and physically distributed architecture. But, as our investigations show, it is unable to cope with hazardous network conditions such as sporadic or recurring packet loss. It either does not detect packet loss or only detects it after long time periods, failing all common network availability targets.

@inproceedings{info3-inproceedings-2018-22, abstract = {The process of monitoring the network, especially for larger ones, is very complex and contains many pitfalls. For instance the balance between granularity of information and their performance impact on the network. With the help of SDN this challenge can become easier, as it offers new methods, mechanisms and opportunities. One of the current most important Open Source controllers is the ONOS SDN controller. According to its developers it is a production ready controller that offers high availability due to its logically centralized and physically distributed architecture. But, as our investigations show, it is unable to cope with hazardous network conditions such as sporadic or recurring packet loss. It either does not detect packet loss or only detects it after long time periods, failing all common network availability targets.}, address = {Hue, Vietnam}, author = {Metter, Christopher and Burger, Valentin and Hu, Zheng and Pei, Ke and Wamser, Florian}, booktitle = {7th IEEE International Conference on Communications and Electronics (IEEE ICCE 2018)}, keywords = {can}, month = {07}, pages = {96 - 101}, title = {Evaluation of the Detection Capabilities of the ONOS SDN Controller}, year = 2018 }

%0 Conference Paper %1 info3-inproceedings-2018-22 %A Metter, Christopher %A Burger, Valentin %A Hu, Zheng %A Pei, Ke %A Wamser, Florian %B 7th IEEE International Conference on Communications and Electronics (IEEE ICCE 2018) %C Hue, Vietnam %D 2018 %P 96 - 101 %T Evaluation of the Detection Capabilities of the ONOS SDN Controller %X The process of monitoring the network, especially for larger ones, is very complex and contains many pitfalls. For instance the balance between granularity of information and their performance impact on the network. With the help of SDN this challenge can become easier, as it offers new methods, mechanisms and opportunities. One of the current most important Open Source controllers is the ONOS SDN controller. According to its developers it is a production ready controller that offers high availability due to its logically centralized and physically distributed architecture. But, as our investigations show, it is unable to cope with hazardous network conditions such as sporadic or recurring packet loss. It either does not detect packet loss or only detects it after long time periods, failing all common network availability targets.
Analytical Model for SDN Signaling Traffic and Flow Table Occupancy and its Application for Various Types of Traffic. Metter, Christopher; Seufert, Michael; Wamser, Florian; Zinner, Thomas; Tran-Gia, Phuoc. In IEEE Transactions on Network and Service Management, 14(3), pp. 603–615. 2017.
- [ Abstract ]
- [ BibTeX ]
- [ EndNote ]
Software Defined Networking (SDN) has emerged as a promising networking paradigm overcoming various drawbacks of current communication networks. The control and data plane of switching devices is decoupled and control functions are centralized at the network controller. In SDN, each new flow introduces additional signaling traffic between the switch and the controller. Based on this traffic, rules are created in the flow table of the switch, which specify the forwarding behavior. To avoid table overflows, unused entries are removed after a predefined time-out period. Given a specific traffic mix, the choice of this time-out period affects the trade-off between signaling rate and table occupancy. As a result, network operators have to adjust this parameter to enable a smooth and efficient network operation. Due to the complexity of this problem caused by the various traffic flows in a network, a suitable abstraction is necessary in order to derive valid parameter values in time. The contribution of this work is threefold. Firstly, we formulate a simple analytical model that allows optimizing the network performance with respect to the table occupancy and the signaling rate. Secondly, we validate the model by means of simulation. Thirdly, we illustrate the impact of the time-out period on the signaling traffic and the flow table occupancy for different data-plane traffic mixes and characteristics. This includes scenarios with single application instances, as well as multiple application instances of different application types in an SDN-enabled network.

@article{info3-article-2017-5, abstract = {Software Defined Networking (SDN) has emerged as a promising networking paradigm overcoming various drawbacks of current communication networks. The control and data plane of switching devices is decoupled and control functions are centralized at the network controller. In SDN, each new flow introduces additional signaling traffic between the switch and the controller. Based on this traffic, rules are created in the flow table of the switch, which specify the forwarding behavior. To avoid table overflows, unused entries are removed after a predefined time-out period. Given a specific traffic mix, the choice of this time-out period affects the trade-off between signaling rate and table occupancy. As a result, network operators have to adjust this parameter to enable a smooth and efficient network operation. Due to the complexity of this problem caused by the various traffic flows in a network, a suitable abstraction is necessary in order to derive valid parameter values in time. The contribution of this work is threefold. Firstly, we formulate a simple analytical model that allows optimizing the network performance with respect to the table occupancy and the signaling rate. Secondly, we validate the model by means of simulation. Thirdly, we illustrate the impact of the time-out period on the signaling traffic and the flow table occupancy for different data-plane traffic mixes and characteristics. This includes scenarios with single application instances, as well as multiple application instances of different application types in an SDN-enabled network.}, author = {Metter, Christopher and Seufert, Michael and Wamser, Florian and Zinner, Thomas and Tran-Gia, Phuoc}, journal = {IEEE Transactions on Network and Service Management}, keywords = {can}, month = {09}, number = 3, pages = {603-615}, title = {Analytical Model for SDN Signaling Traffic and Flow Table Occupancy and its Application for Various Types of Traffic}, volume = 14, year = 2017 }

%0 Journal Article %1 info3-article-2017-5 %A Metter, Christopher %A Seufert, Michael %A Wamser, Florian %A Zinner, Thomas %A Tran-Gia, Phuoc %D 2017 %J IEEE Transactions on Network and Service Management %N 3 %P 603-615 %T Analytical Model for SDN Signaling Traffic and Flow Table Occupancy and its Application for Various Types of Traffic %V 14 %X Software Defined Networking (SDN) has emerged as a promising networking paradigm overcoming various drawbacks of current communication networks. The control and data plane of switching devices is decoupled and control functions are centralized at the network controller. In SDN, each new flow introduces additional signaling traffic between the switch and the controller. Based on this traffic, rules are created in the flow table of the switch, which specify the forwarding behavior. To avoid table overflows, unused entries are removed after a predefined time-out period. Given a specific traffic mix, the choice of this time-out period affects the trade-off between signaling rate and table occupancy. As a result, network operators have to adjust this parameter to enable a smooth and efficient network operation. Due to the complexity of this problem caused by the various traffic flows in a network, a suitable abstraction is necessary in order to derive valid parameter values in time. The contribution of this work is threefold. Firstly, we formulate a simple analytical model that allows optimizing the network performance with respect to the table occupancy and the signaling rate. Secondly, we validate the model by means of simulation. Thirdly, we illustrate the impact of the time-out period on the signaling traffic and the flow table occupancy for different data-plane traffic mixes and characteristics. This includes scenarios with single application instances, as well as multiple application instances of different application types in an SDN-enabled network.
Analytic Model for SDN Controller Traffic and Switch Table Occupancy. Metter, Christopher; Seufert, Michael; Wamser, Florian; Zinner, Thomas; Tran-Gia, Phuoc. In 12th International Conference on Network and Service Management (CNSM). Best Paper Award, Montreal, Canada, 2016.
- [ Abstract ]
- [ BibTeX ]
- [ EndNote ]
Software Defined Networking (SDN) is a major paradigm in the field of current communication networks. SDN is used as the basis of many new networks although few performance models are available in the literature, and the majority of performance evaluations are based primarily on practical measurements. To fill this gap, we develop an analytical model to assess SDN control plane traffic as well as the occupancy of the flow table of an SDN switch. The contribution of this work is the formulation of the model for the performance-decisive parameters control-plane traffic and flow table occupancy and the application of the model for different data plane traffic characteristics. In the end, there is a discussion about the setting of time-out values for storing flow entries in the switch flow table depending on the traffic characteristics in the data plane. The trade-off between the signaling traffic in the control plane and the occupancy of the flow table is discussed in order to minimize both.

@inproceedings{info3-inproceedings-2016-17, abstract = {Software Defined Networking (SDN) is a major paradigm in the field of current communication networks. SDN is used as the basis of many new networks although few performance models are available in the literature, and the majority of performance evaluations are based primarily on practical measurements. To fill this gap, we develop an analytical model to assess SDN control plane traffic as well as the occupancy of the flow table of an SDN switch. The contribution of this work is the formulation of the model for the performance-decisive parameters control-plane traffic and flow table occupancy and the application of the model for different data plane traffic characteristics. In the end, there is a discussion about the setting of time-out values for storing flow entries in the switch flow table depending on the traffic characteristics in the data plane. The trade-off between the signaling traffic in the control plane and the occupancy of the flow table is discussed in order to minimize both.}, address = {Best Paper Award, Montreal, Canada}, author = {Metter, Christopher and Seufert, Michael and Wamser, Florian and Zinner, Thomas and Tran-Gia, Phuoc}, booktitle = {12th International Conference on Network and Service Management (CNSM)}, keywords = {can}, month = 10, title = {Analytic Model for SDN Controller Traffic and Switch Table Occupancy}, year = 2016 }

%0 Conference Paper %1 info3-inproceedings-2016-17 %A Metter, Christopher %A Seufert, Michael %A Wamser, Florian %A Zinner, Thomas %A Tran-Gia, Phuoc %B 12th International Conference on Network and Service Management (CNSM) %C Best Paper Award, Montreal, Canada %D 2016 %T Analytic Model for SDN Controller Traffic and Switch Table Occupancy %X Software Defined Networking (SDN) is a major paradigm in the field of current communication networks. SDN is used as the basis of many new networks although few performance models are available in the literature, and the majority of performance evaluations are based primarily on practical measurements. To fill this gap, we develop an analytical model to assess SDN control plane traffic as well as the occupancy of the flow table of an SDN switch. The contribution of this work is the formulation of the model for the performance-decisive parameters control-plane traffic and flow table occupancy and the application of the model for different data plane traffic characteristics. In the end, there is a discussion about the setting of time-out values for storing flow entries in the switch flow table depending on the traffic characteristics in the data plane. The trade-off between the signaling traffic in the control plane and the occupancy of the flow table is discussed in order to minimize both.

Student projects and theses

Analytical Model for SDN Signaling Traffic and Flow Table Occupancy and its Application for Various Types of Traffic. Metter, Christopher; Seufert, Michael; Wamser, Florian; Zinner, Thomas; Tran-Gia, Phuoc. In IEEE Transactions on Network and Service Management, 14(3), pp. 603–615. 2017.
- [ Abstract ]
- [ BibTeX ]
- [ EndNote ]
Software Defined Networking (SDN) has emerged as a promising networking paradigm overcoming various drawbacks of current communication networks. The control and data plane of switching devices is decoupled and control functions are centralized at the network controller. In SDN, each new flow introduces additional signaling traffic between the switch and the controller. Based on this traffic, rules are created in the flow table of the switch, which specify the forwarding behavior. To avoid table overflows, unused entries are removed after a predefined time-out period. Given a specific traffic mix, the choice of this time-out period affects the trade-off between signaling rate and table occupancy. As a result, network operators have to adjust this parameter to enable a smooth and efficient network operation. Due to the complexity of this problem caused by the various traffic flows in a network, a suitable abstraction is necessary in order to derive valid parameter values in time. The contribution of this work is threefold. Firstly, we formulate a simple analytical model that allows optimizing the network performance with respect to the table occupancy and the signaling rate. Secondly, we validate the model by means of simulation. Thirdly, we illustrate the impact of the time-out period on the signaling traffic and the flow table occupancy for different data-plane traffic mixes and characteristics. This includes scenarios with single application instances, as well as multiple application instances of different application types in an SDN-enabled network.

@article{info3-article-2017-5, abstract = {Software Defined Networking (SDN) has emerged as a promising networking paradigm overcoming various drawbacks of current communication networks. The control and data plane of switching devices is decoupled and control functions are centralized at the network controller. In SDN, each new flow introduces additional signaling traffic between the switch and the controller. Based on this traffic, rules are created in the flow table of the switch, which specify the forwarding behavior. To avoid table overflows, unused entries are removed after a predefined time-out period. Given a specific traffic mix, the choice of this time-out period affects the trade-off between signaling rate and table occupancy. As a result, network operators have to adjust this parameter to enable a smooth and efficient network operation. Due to the complexity of this problem caused by the various traffic flows in a network, a suitable abstraction is necessary in order to derive valid parameter values in time. The contribution of this work is threefold. Firstly, we formulate a simple analytical model that allows optimizing the network performance with respect to the table occupancy and the signaling rate. Secondly, we validate the model by means of simulation. Thirdly, we illustrate the impact of the time-out period on the signaling traffic and the flow table occupancy for different data-plane traffic mixes and characteristics. This includes scenarios with single application instances, as well as multiple application instances of different application types in an SDN-enabled network.}, author = {Metter, Christopher and Seufert, Michael and Wamser, Florian and Zinner, Thomas and Tran-Gia, Phuoc}, journal = {IEEE Transactions on Network and Service Management}, keywords = {can}, month = {09}, number = 3, pages = {603-615}, title = {Analytical Model for SDN Signaling Traffic and Flow Table Occupancy and its Application for Various Types of Traffic}, volume = 14, year = 2017 }

%0 Journal Article %1 info3-article-2017-5 %A Metter, Christopher %A Seufert, Michael %A Wamser, Florian %A Zinner, Thomas %A Tran-Gia, Phuoc %D 2017 %J IEEE Transactions on Network and Service Management %N 3 %P 603-615 %T Analytical Model for SDN Signaling Traffic and Flow Table Occupancy and its Application for Various Types of Traffic %V 14 %X Software Defined Networking (SDN) has emerged as a promising networking paradigm overcoming various drawbacks of current communication networks. The control and data plane of switching devices is decoupled and control functions are centralized at the network controller. In SDN, each new flow introduces additional signaling traffic between the switch and the controller. Based on this traffic, rules are created in the flow table of the switch, which specify the forwarding behavior. To avoid table overflows, unused entries are removed after a predefined time-out period. Given a specific traffic mix, the choice of this time-out period affects the trade-off between signaling rate and table occupancy. As a result, network operators have to adjust this parameter to enable a smooth and efficient network operation. Due to the complexity of this problem caused by the various traffic flows in a network, a suitable abstraction is necessary in order to derive valid parameter values in time. The contribution of this work is threefold. Firstly, we formulate a simple analytical model that allows optimizing the network performance with respect to the table occupancy and the signaling rate. Secondly, we validate the model by means of simulation. Thirdly, we illustrate the impact of the time-out period on the signaling traffic and the flow table occupancy for different data-plane traffic mixes and characteristics. This includes scenarios with single application instances, as well as multiple application instances of different application types in an SDN-enabled network.

Networking and SDN