Software defined networking (SDN) and flexible grid optical transport technology are two key technologies that allow network operators to customize their infrastructure based on application requirements and therefore minimizing the extra capital and operational costs required for hosting new applications. In this paper, for the first time we report on design, implementation & demonstration of a novel OpenFlow based SDN unified control plane allowing seamless operation across heterogeneous state-of-the-art optical and packet transport domains. We verify and experimentally evaluate OpenFlow protocol extensions for flexible DWDM grid transport technology along with its integration with fixed DWDM grid and layer-2 packet switching.
©2013 Optical Society of America
Future and emerging Internet applications are becoming cloud-based applications with a common requirement for a high capacity optical network infrastructure. This requires end-to-end high capacity connectivity between users and geographically distributed remote data centers traversing multiple network domains. Therefore a challenge for network operators is the efficient end-to-end delivery of application traffic in terms of network operation, control and management when traversing heterogeneous transport technologies. Furthermore, each application type requires its own specific network services to satisfy its unique bandwidth, quality of service (QoS) and dynamicity requirements. The current technical and operational complexities as well as CAPEX and OPEX considerations limit the ability of network operators to setup and configure dedicated optical network for each application type in a scalable manner.
Software defined networking (SDN) and emerging optical transport technologies such as flexi WDM grid are key technology enabler for addressing aforementioned challenges. SDN allows programmability of network functions and protocols by decoupling the data plane and the control plane, which are vertically integrated in many networking equipment’s . Central to the SDN technology is a logically centralized controller hosting a network operating system. It abstracts the underlying transport and switching technology while constructing and presenting a logical map of the entire network to services or control applications implemented on top of it. The SDN technology allows network operators to manipulate logical map of the network and create multiple co-existing network slices (virtual networks) in a technology and protocol agnostic manner. Furthermore, the separation of control plane and data plane supported by a logically centralized control plane makes it a suitable candidate for an integrated control plane supporting multiple domain and multiple transport technologies.
OpenFlow (OF) is an open, vendor and technology agnostic protocol that allows separation of data and control plane and therefore is a suitable candidate for realization of SDN. It is based on flow switching with the capability to execute software/user defined routing, control and management applications in a controller (i.e. OF controller) outside of the data path. In an OF-based optical network, a centralized controller utilizes the OF protocol to directly and securely manipulate the forwarding plane of the network elements (NEs) (e.g., optical switches). OpenFlow provides a network control framework with a common hardware abstraction. Using these resource abstractions, the centralized controller is able to configure the state of the switches. Although the current OF protocol is originated from the packet switching (i.e. Ethernet), the OF based SDN control framework and extensions of OF protocols can be considered as a realistic unified control plane solution for integration of packet and optical circuit switched networks .
A wave of new innovations & technologies has been introduced to optical networking in recent years. One of the most promising technologies, providing flexibility in the rigid optical network is Flexible Grid DWDM technology. It allows allocation of an arbitrary and appropriate spectrum range, as described in G.694.1 , and modulation format to an optical path according to application bandwidth and QoS requirements taking into account optical physical layer attributes such as impairments.
In order to utilize this flexibility, bandwidth/bit-rate granularity and scalability offered by flexible grid, a dynamic and intelligent control & management plane integrating with existing fixed DWDM grid and packet networks is required. Software defined networking using OpenFlow abstractions, that includes flexible grid, fixed grid & packet networks, enables network operator to efficiently and dynamically integrate different transport technology domains i.e. from packet based campus to traditional fixed WDM core as well as more advanced flexi WDM core domains. In addition it allows them to reconfigure their multi-domain heterogeneous transport network, deploy additional capacity, and provide greater bandwidth elasticity so as to support new services in an on demand manner based on applications requirements, with a minimum amount of manual intervention, hardware deployment and engineering complexity.
This paper introduces a novel unified control plane approach for multi-domain multi-transport networks based on SDN framework with OpenFlow as a protocol enabler. For the first time we report on extensions to existing OF protocol specification to support heterogeneous network comprising of flexible and fixed DWDM grid transport technologies integrated with layer-2 packet networks. Furthermore, we experimentally demonstrate the proposed solution over an International test-bed comprising of different domains from different geographical locations (UK, Spain & Japan) using a novel SDN application for multi-domain network slicing. To the best of authors’ knowledge, it is for the first time that a field trial comprising flexi, fixed grid optical networks and packet network controlled by an OpenFlow based control plane has been reported. The previous OF-based field trials only considered fixed-grid optical networks .
The remainder of this paper is organized as follows. Section 2 describes the proposed SDN/OF architecture and its building blocks. It details the OF extensions to support fixed and flexible grid optical network elements and also verifies co-ordination of OF with existing GMPLS control plane over commercial fixed grid network. Furthermore a novel SDN based application is introduced which utilizes the proposed OF protocol extension to provide virtual network slices across multiple domains. Finally section 3 describes experimental demonstration and evaluation of the proposed solution over a large international test-bed.
2. A novel OpenFlow based unified control plane for multi-domain packet over fixed/flexible grid optical networking
Figure 1(a) shows the architectural block diagram of the proposed OF based unified SDN control plane. Central to the proposed architecture is extensions to the OF protocol to support fixed & flexible grids optical DWDM network as well as the multi-domain operation. OF enabled devices capabilities are abstracted using OF protocol and controller uses these abstracted information to build a technology and domain specific topology database. Using the knowledge of different network resources and their associated technology constraints described in the topology database, the OF controller can facilitate an SDN application like a Path Computation Element (PCE) or virtual optical networks provider (network slicer) to provide a path or network slice across different technology & administrative domains.
To realize such an SDN/OpenFlow based multi-domain control plane, the following building blocks were implemented
- 1. Extend OF protocol to include flexible and fixed DWDM grid capability
- 2. Extend OF to support optical functionalities like switching constraints, power equalization, impairments
- 3. Extend OF for multi-domain/multi-technology operation such as peering and flow mapping between multiple domains and/or technologies
- 4. SDN application to utilize the abstracted resources provided by OF controller. Following section describes in more detail the mentioned development blocks.
2.1 Openflow protocol extension
In an OpenFlow controlled network, forwarding functions are performed in the OF switch according to the flow table entries and the controller is responsible for hosting network applications (routing, path computation etc.) that can use abstracted information provided by OF protocol to define and enforce control plane decision by inserting flow switching rules or action into the OF switch. The OF protocol comprises of three main messages: Switch Feature, Flow_Mod and CPort_status messages. Switch Feature messages are used by OF-enabled nodes to describe network element capabilities and constraints while the Flow_Mod messages are used by controller to insert flows and define flow switching actions on the OF nodes. CPort_status describes the port change information e.g if the link is down or if the bandwidth on the link is updated. Since currently there aren’t any optical equipment vendors supporting OF, hence we propose a generic OF agent to be placed on the optical node to enable it to support OF protocol. As shown in Fig. 1(b), it is a modular OpenFlow agent capable of supporting OF protocol providing hardware abstractions for the controller via an OpenFlow interface. This agent utilizes the NE’s management interface (i.e., e.g. SNMP, Vendor API) to communicate with the data plane. A generic and novel Resource Model is designed & implemented to maintain NE’s configurations (wavelengths, port capabilities & switching constraints). The Resource Model deals with the complexity of the NEs capabilities and represents NE capability and constraints to the controller in a generic format. It also holds the flow tables and also the mapping required for integration with other domains. For e.g. if its packet to optical flow map then the resource model at the optical node will not only hold the wavelength supported by the optical node but also provide the wavelengths supported to reach the packet peer which is discovered by the discovery protocol. OF agent also includes the OF channel, which is responsible for communication with the extended OF controller.
While the existing OF circuit switch addendum V0.3  extends OF from the packet domain to TDM circuits, it does not support flexible DWDM grid technology. Furthermore, the current OF protocol specification as well as OF controller reference implementations (e.g. Floodlight or NOX)  do not include optical functionalities like power equalization, impairments etc. and also there is no support for multi-domain/multi-technology operation such as peering and flow mapping between multiple domains and/or technologies. This section describes extensions to the OF protocol and the design of a NOX based OF controller to extend the circuit specification to support flexible grid functionalities and multi-domain operation for seamless path computation between different network domains.
2.2 Flexible grid & multi-domain
Traditional fixed grid network uses fixed size optical spectrum frequency ranges with channel spacing of 100 or 50GHz while in flexible grid optical networks both channel spacing and channel bandwidth are variable and flexible.
Figure 2(a) shows the flow specification for packet, fixed and flexible grid optical networks. For the packet domain a flow can be any combination of the L2-L4 header, for fixed grid optical network a flow can be identified by a fixed flow identifier comprising port, wavelength and signal type fields associated with a specific optical switch. In a flexible grid optical network a flow is identified by a flexible flow identifier comprising port, central frequency (CF), frequency slot bandwidth (FSB) and type of signal. Using these common flow identifiers an OF node can perform actions like cross connect ports, modify address, port/tag translation etc using Flow_Mod messages.
The ITUT G.694.1 document, describes a new optical spectrum management specification for Flexible Grid DWDM optical networks. It defines nominal frequency of flexible DWDM grids by equation 193.1 + n × 0.00625 (THz) which is used to calculate the central frequency of a frequency slot and equation 12.5 GHz × m yields the slot width, where n is an integer and m is a positive integer. The permissible granularity of m and n for flexible grid equipment can be determined using OFs Switch_Feature messages. The controller builds topology utilising information from the Switch_Feature message and then in-order to control the flexible grid optical node CFlow_Mod (circuit Flow_Mod) messages which includes m and n values are exchanged between the controller and node.
To implement the proposed multi domain, multi technology control plane architecture the Switch_Feature, CFlow_Mod and CPort_Status messages of the OF protocol were extended to support different types of optical NEs: fixed grid optical cross connect (WDM OXC), bandwidth variable optical cross connect (BV OXC) and bandwidth variable optical transponder (BVT). The Switch_Feature message extension supports optical NE capabilities including: central frequency, spectrum range, and granularity of BVT and BV OXC; number of ports and wavelength channels of WDM OXC; switching constraints of the WDM OXC including peering connectivity inside and across multiple domains
For multi-domain aspects the controller is made aware of domain constraints by utilizing intra-domain and inter-domain flow tables. An intra–domain flow table holds flow identifiers and associated actions for each network element within a particular domain. For optical NEs, an action is defined as a cross connection associated with one or more flow identifiers. For a flexible grid NE, the action also includes CF and FSB. Inter-domain flow tables hold flow identifiers and associated actions for NEs that interconnect between neighbouring domains. Actions in flow tables that are associated with two heterogeneous technology domains must comply with multi-domain mapping rules described in Fig. 2(b).
2.3 Fixed DWDM grid extension & integration with GMPLS
OF extensions to support fixed grid WDM optical network functions are already defined in the OF circuit specification . However a SDN enabled control plane based on OF requires sophisticated optical functionalities like power equalization, impairments etc. along with a set of network control applications such as path computation for control and anagement of the network. Currently, generalized multi protocol label switching (GMPLS) is the most comprehensive control protocol suit that includes all the required functionalities and mechanisms for control and management of the optical network. But GMPLS has failed to attract provider’s interest owing to its complex, inflexible and closed architecture . In this paper, derived from  , we propose and evaluate a hybrid OF and GMPLS control plane approach to take advantages of the rich network control functionalities of GMPLS as well as openness and flexibility of OF. Furthermore, we experimentally evaluate the performance of an extended Optical OF deployment (i.e., OF agents, controllers along with a sample network application) in a converged packet and fixed-grid optical network using both integrated OF-GMPLS and a pure standalone extended OF approaches using commercial reconfigurable optical add drop multiplexers (ADVA ROADMs).
In hybrid GMPLS-OpenFlow approach, each NE functions as an OF enabled switch Fig. 3(a) . The OF controller receives information regarding the topology and resources using extended OF protocol and agents. However, path computation, lightpath establishment and teardown are performed utilizing the GMPLS control plane. Extended OF agents, controller and sample network control application are developed considering loose and explicit lightpath establishment. In the former case only ingress and egress NEs and ports are specified and the control plane handles the path computation and establishment. In the latter case, the controller is able to specify the full details of the lightpath (i.e., all switches and ports along the lightpath) and the control plane verifies the feasibility of the lightpath and performs the establishment. The main idea is to provide an extended OF interface to the OF controller for reusing the GMPLS functionalities. The extended OF controller exploits the GMPLS library to compute, establish and verify lightpaths using built-in Path Computation Element, lightpath establishment/teardown, switching constraints, and power equalization functionalities.
We also propose extended pure OF approach Fig. 3(b), which includes an OF agent on each NE. Our developed agent utilizes an internal API to communicate with the NEs, and the extended OF protocol to communicate with the developed OF controller. In this approach the controller utilizes the switch feature requests/replies (e.g., ports, wavelength) to construct the network topology and CFlow_MOD (flow specification) message to control the cross connections in the optical switches. Switching constraints are sent to the controller using a generic vendor extensions messages. The CFlow_MOD message is used to establish or teardown the cross-connections in each NE. A specialized path computation network application (OF_PCE) is implemented and integrated in the extended controller. It is responsible to compute a lightpath inside the optical domain with proper consideration for switching constraints.
Both extensions rely upon OF agent software, which provides a novel hybrid switch abstraction to the OF controller beyond V0.3 circuit addendum. The optical agents extend the existing OF circuit extension by introducing switching constraint & power equalization messages. The agents resource model calculates the constraints from the capabilities information and exchanges it via OF messages using the OF channel. Similarly the power equalization OF message is injected upon cross connection initiation via the OF controller.
2.4 SDN application for virtual network slicing over mixed fixed & flexi-grid domains
As an SDN application, to validate the proposed control plane, we have developed an algorithm bundle, including several algorithms designed for different network slicing scenarios (i.e. slicing over single/ multiple fixed-grid network domains, single/multiple flexi-grid network domains, mixed fixed- & flexi-grid network domains), which is running on top of the OF controller to compose virtual network (VN) slices over flexi and fixed-grid domains. The algorithm supports two main functionalities one is to calculate the best path from source to destination for all connectivity within each VN slice and the latter is to find the optimum spectrum across domains to fulfil bandwidth requirements of each VN slice request. The algorithm bundle reads the information of physical networks and the user requests from the OF controller. The physical network information obtained from the topology database of the OF controller not only involves the nodes and their connectivity’s but also the domain constraints and physical layer impairments. Utilizing the flow mapping description at Fig. 2(b) the application can serve requests taking into consideration the domain constraints like supported wavelengths, impairments etc.
In each VN request, the VN topology, virtual resources requirements (e.g. bandwidth), the virtual network quality (e.g. latency), and the VN holding time are specified. The information of physical networks indicates the network topology and the availability of physical resources (i.e. ports, wavelength channels, and spectrum slots) of each domain, and the inter-domain connectivity, etc. Taking into account the above inputs and optical constraints (i.e. wavelength continuity and spectrum continuity, impairments, etc.), the algorithm bundle will compose a VN slice that can satisfy the user’s request and minimize the utilized physical resources.
After a request arrives, as depicted in Fig. 4(a) , the algorithm bundle will first determine whether the VN slicing request is a single domain or multi-domain problem. Then according to the feature of each involved domain (i.e. fixed-grid or flexi-grid), the corresponding algorithm is adopted. If it’s a flexible-grid request then the path is calculated using the Routing and Spectrum Assignment algorithm and if it’s a fixed-grid domain then Routing and Wavelength Assignment algorithm is chosen.
The hop-count shortest path routing algorithm is used to find a physical path for each virtual link to reduce the involved NEs and/or domains (for multi-domain). Figure 4(b) shows the bandwidth request being resolved in both fixed and flexible grid network domains. In fixed-grid domain, the first-fit wavelength assignment is used afterwards to allocate the required channels. In flexi-grid domain, among all the available spectrum slots, the one that can have the minimum residual spectrum after the spectrum assignment for the requested bandwidth is chosen, which can mitigate the spectrum fragmentation effect and save wide available spectrum slot for future requests. The wavelength continuity and/or spectrum continuity needs to be satisfied in the case of the absence of wavelength conversion capability. In the case where the spectrum continuity from fixed grid domain to flexi grid is limited due to the fixed spectrum the application will check and accept the request depending on the spectrum availability on the flexi grid. After running the algorithm bundle, all the involved NEs, the port no., the channel no., the m and n values for configuring the flexi-grid equipment are generated as outputs and send to the OF controller.
3. Experimental setup & demonstration
To experimentally evaluate the proposed architecture and OF extensions we built a heterogeneous multi-domain international test-bed comprising flexible and fixed grid optical domains along with layer 2 packet switched domain in High Performance Networks group at University of Bristol (UnivBris (formerly in university of Essex), UK), a flexible grid and fixed grid optical network domain in KDDI Labs (Japan) and a fixed grid optical network domain in CTTC (Spain) as shown in Fig. 4. The UnivBris fixed-flexible grid test-bed is comprised of an in-house built 8x8 (4x4 bidirectional) BV OXC utilizing two BV WSS switches with internal recirculation fibre loops to emulate multiple nodes; a BV transponder (BV TX & BV RX) supporting C-band and 3 ADVA FSP3000 ROADMs with two active wavelength channels. The UnivBris packet switched test-bed comprises 4 NEC IPX and Extreme Summit OF enabled GE switches. The CTTC fixed grid test-bed includes 4 ROADMs in mesh topology supporting 6 wavelengths. The KDDI flexible-fixed grid test-bed includes 4 ROADM in a mesh topology supporting 3 wavelengths, 3 hardware-emulated BV OXCs and one emulated BVT. The CTTC test-bed is connected to UnivBris flexible-grid and KDDI fixed-grid test-bed. The KDDI flexible-grid test-bed is connected to UnivBris flexible-grid test-bed to form a ring among the domains. The extended NOX based OF controller is hosted in UnivBris. The International control and data plane connectivity between test-beds is emulated over VPN Tunnels over internet.
In order to evaluate the fixed grid OF extensions for both the integrated GMPLS-OF and pure OF approach, the UnivBris Islands ADVA ROADM and packet switches are used. The centre part of the overall testbed at Fig. 4 depicting the UnivBris,UK testbed was used for the integrated approach and the results are depicted at Fig. 6 . Figure 6(a) shows the blocking rate versus the load result of hybrid (explicit, loose path) and pure OF approaches were 23%, 23%, and 22% respectively. Lightpath requests are generated according to a Poisson process and uniformly distributed among all node pairs. Both inter-arrival of request and their holding times are exponentially distributed. Imposed load to the extended controller in terms of lightpath requests (100 requests) are varied from 50 to 300 Erlangs. The high blocking rate is mainly due to the limited number of client ports per each NE. Next in Fig. 6(b) the End-to-End lightpath setup time (required time for cross connection along the path) for both approaches were calculated using the experimental setup. The individual network element setup times categorised based on hardware, power equalization & teardown times. OF approach was better, owing to its ability to equalize power parallel on involved network elements. In order to verify the OF agent performance, hardware setup time for different load characteristics are noted at Fig. 6(c) We observed that due to lightweight path setup procedure in Pure OF approach, faster setup times are achieved. Performance of extended OF controllers (i.e., OpenVSwitch (OVS) and NOX-based) in terms of flow operation per second for different number of OF enabled switches is depicted in Fig. 6(d). The OVS-based extended OF controller demonstrates higher performance than NOX-based one for a single threaded application.
With the fixed grid approaches mentioned above we implemented the proposed flexible grid OF extensions to evaluate the performance of the proposed architecture over an international experimental test-bed. We use the aforementioned network slicing SDN application that composes/calculates virtual topology (network slices) comprising multi-domain multi-technology optical paths based on request from a user (here random topology). Using the network slicing application three different multi-domain network slices was created with different bandwidth. To reduce complexity of the experiment, same bandwidth is assigned to all connectivity within a slice. Optical spectrum connectivity of the three network slices and their details are shown in Fig. 7(a) and the domains involved for corresponding slices is shown in Fig. 7(b).
We also measured the end-to-end path setup time as well as timing performance of the OF controller with different NEs for different network scenarios i.e. different slice topologies in Fig. 7(c). These results were used to verify the functionalities of the proposed control plane and OF extensions for the physical path setup in three different configurations. The pathsetup time is affected significantly by control plane VPN delay as well as the communication method between each OF agents and its corresponding NE. We used compressed video streaming for multi domain test and 4K HD uncompressed video streaming for single domain and observed error free transmissions, while monitoring the status of involved NEs and bandwidth spectrum of each network slice.
Figure 8(a) shows a snap shot of OF message exchange between NE OF agents in different domains and OF controller. Figure 8(a)-top shows OF feature messages exchanges for CTTC, KDDI and UnivBris domain. Figure 8(a)-middle, shows CFlow_Mod between UnivBris and CTTC domain to configure fixed and flexi grid devices. Figure 8(b)-bottom shows the center frequency and bandwidth (“m” and “n” values) exchange between controller and BV OXC agent.
This flow of OF messages exchange verifies the circuit switching extensions for multi-technology over geographically distributed multi-domains. On the other hand, in Fig. 8(b) we show the performance of the various operational parts of the OF controller. The controller setup time indicates the time required for creating and processing OF messages in both OF controller and agents. The hardware setup time includes the controller setup time and the time taken for each agent to configure its corresponding NE upon receiving CFlow_Mod message. The algorithm time is the time for the SDN application to compute a network slice. Figure 8(b) includes the above timing components as well as the time required for the controller to exchange messages (across the control plane VPN) with the different agents in the different domains depicting satisfying timings for optical control plane operation.
In this work we successfully implemented OF-based network connectivity provisioning and network slicing in an international test-bed across multiple optical transport technologies and geographical domains. Demonstrated & evaluated a SDN/OF based unified control plane supporting fixed, flexible grid optical transport and packet networks. Also evaluated the performance and feasibility of hybrid GMPLS-OF and pure extended OF for control and management of optical circuit switched networks.
We can summarize from the results that integrated GMPLS-OF solution provides satisfactory performance and it would be advantageous to use GMPLS’s well-defined extensions for control of optical network. We also demonstrated an SDN application capable of creating virtual slices from a heterogeneous (fixed, flexi & packet) physical substrate thereby validating the multi-technology multi-domain OF based control plane performance. Future work will bring in network virtualization via an optical flowvisor . This will allow infrastructure providers to provide virtual networks taking into account domain and technology policies and also open new avenues for better network utilization.
Our experiments demonstrate that OF based SDN can provide an extensible control framework for packet over optical transport embracing existing and emerging optical transport technologies. The work pioneers new features to the OF circuit specifications and aims to enable dynamic, flexible networking as well as virtualization features.
This work is partially supported by EU funded projects OFELIA, FIBRE and UK funded EPSRC PATRON, Hyper Highway and funded by Spanish Ministry of Science and Innovation (MICINN) through DORADO project (TEC2009-07995).
References and links
1. N. McKeown, T. Anderson, H. Balakrishnan, G. Parulkar, L. Peterson, J. Rexford, S. Shenker, and J. Turner, “OpenFlow: enabling innovation in campus networks,” ACM SIGCOMM Comput. Commun. Rev. 38(2), 69–74 (2008). [CrossRef]
2. S. Das, G. Parulkar, N. McKeown, P. Singh, D. Getachew, and L. Ong, “Packet and circuit network convergence with OpenFlow, ” in Optical Fiber Communication Conference and Exposition and National Fiber Optic Engineers Conference (OFC/NFOEC 2010), Technical Digest (CD) (Optical Society of America, 2010), paper OTuG1. http://www.opticsinfobase.org/abstract.cfm?URI=OFC-2010-OTuG1. ONF white paper, “Software-Defined Networking: The New Norm for Networks,” (ONF 2012).
3. ITU-T Recommendation G.694.1, Feb 2012.
4. L. Liu, D. Zhang, T. Tsuritani, R. Vilalta, R. Casellas, L. Hong, I. Morita, H. Guo, J. Wu, R. Martínez, and R. Muñoz, “First field trial of an OpenFlow-based unified control plane for multi-layer multi-granularity optical networks, ” in Optical Fiber Communication Conference and Exposition and National Fiber Optic Engineers Conference (OFC/NFOEC 2012), Technical Digest (CD) (Optical Society of America, 2012), paper PDP5D.2.
5. S. Das, “Extensions to the OF Protocol in support of Circuit Switching, “addendum v0.3, June 2010.
6. N. Gude, T. Koponen, J. Pettit, B. Pfaff, M. Casado, N. McKeown, and S. Shenker, “NOX: Towards an Operating System for Networks,” ACM SIGCOMM Comput. Commun. Rev. 38(3), 105–110 (2008). [CrossRef]
7. S. Das, G. Parulkar, and N. McKeown, “Why OpenFlow/SDN can succeed where GMPLS failed,” in European Conference and Exhibition on Optical Communication, OSA Technical Digest (online) (Optical Society of America, 2012), paper Tu.1.D.1.
8. S. Azodolmolky, R. Nejabati, E. Escalona, R. Jayakumar, N. Efstathiou, and D. Simeonidou, “Integrated OpenFlow-GMPLS control plane: an overlay model for software defined packet over optical networks,” Opt. Express 19(26), B421–B428 (2011). [CrossRef] [PubMed]
9. M. Channegowda, P. Kostecki, N. Efstathiou, S. Azodolmolky, R. Nejabati, P. Kaczmarek, A. Autenrieth, J. P. Elbers, and D. Simeonidou, “Experimental evaluation of extended OpenFlow deployment for high-performance optical networks,” in Proceedings of ECOC, Tu.1.D.2 (2012).
10. S. Azodolmolky, R. Nejabati, S. Peng, A. Hammad, M. P. Channegowda, N. Efstathiou, A. Autenrieth, P. Kaczmarek, and D. Simeonidou, “Optical FlowVisor: an OpenFlow-based optical network virtualization approach,” in Optical Fiber Communication Conference, OSA Technical Digest (Optical Society of America, 2012), paper JTh2A.41.