Performance of Wavelength Routed Optical Networks Employing GMPLS-based Control Plane Fuead Ali
Mohd Nazri Ismail
Zuraini Zainol
Universiti Kuala Lumpur 1016, Jalan Sultan Ismail Kuala Lumpur, Malaysia
[email protected]
National Defense University Malaysia Kem Sungai Besi Kuala Lumpur, Malaysia
[email protected]
National Defense University Malaysia Kem Sungai Besi Kuala Lumpur, Malaysia
[email protected]
KEYWORDS
guaranteed QoS. Therefore, it is important for network service providers to maximize capacity utilization from their network infrastructures, as is equally important to provide the QoS guarantees for different type of users. One of the main areas of interest in optical networks technology revolves on the automated service delivery for users [1]. In this optical domain, this mechanism is simply defined as the provisioning of the light paths; which is also known as the optical connections. This process must be dynamic, as implemented in the IP domain for the request for connectivity between IP hosts. As such, the study in this research attempts to quantitatively analyze the performance of such network in a dynamic optical connections provisioning process. In this study, the interest in on the all optical network architecture which employs the wavelength routing scheme. The rest of the paper is organized as follows; Section 2 presents the research motivation. The model implementation and characteristics are described in Section 3. Section 4 presents the simulation experiments accompanied with some results. Section 5 concludes this paper and gives some insights on future works.
GMPLS, control plane, performance, optical network, modeling, simulation
2
ABSTRACT proliferation 1
The of optical network technology has helped to alleviate the emerging of next generation transport networks; also known as intelligent optical network. Network functions previously processed in the electrical domain have now gradually migrated to the optical domain, thus giving a huge upgrade in terms of the network capacity. However, as these networks carry a huge volume of traffic, it is imperative that the performance of these networks is measured precisely. This article analyses such a network from modeling perspective which has been developed and simulated using Omnet++ Simulation platform. The GMPLS based optical control plane has been used entirely to build such a network model.
CCS CONCEPTS • Networks → Network performance evaluation; Network performance modeling; Network simulations
ACM Reference format: F. Ali, M.N. Ismail and Z. Zainol. 2018. Performance of Wavelength Routed Optical Networks Employing GMPLS-based Control Plane. In Proceedings of 12th International Conference on Ubiquitous Information Management and Communication, Langkawi, Malaysia, January 2018, (IMCOM’18), 45 pages. DOI: https://doi.org/10.1145/3164541.3164550
1
INTRODUCTION
As the demand for network capacity in the Internet increases, the optical transport network must be more responsive and adaptive in order the drive towards service-oriented networks. This requires optical connections to be managed dynamically, coupled with © 2018 Association for Computing Machinery. ACM acknowledges that this contribution was authored or co-authored by an employee, contractor or affiliate of a national government. As such, the Government retains a nonexclusive, royalty-free right to publish or reproduce this article, or to allow others to do so, for Government purposes only. IMCOM '18, January 5–7, 2018, Langkawi, Malaysia © 2018 Association for Computing Machinery. ACM ISBN 978-1-4503-6385-3/18/01…$15.00 https://doi.org/10.1145/3164541.3164550
RESEARCH MOTIVATION
The current exponential growth of demands for internet data traffic presents a huge challenge for most network service providers. In order to stay competitive, network providers must be able to meet these demands while keeping the operation and running cost at a considerable level. As a result, deployment for a new network infrastructure seems prohibitive since it requires enormous amount of capital investments. In contrast, utilizing the current network infrastructure coupled with strategies on varying and increasing the service offerings can be an interesting option. The advent of Generalized Multiprotocol Label Switching (GMPLS) [2] as optical control plane has helped transforms the future optical network technology. From being a merely point-topoint connection, optical network has reached an advanced state where a number of intelligent functions are ready to be incorporated to the optical domain, such as routing, topology discovery and signaling. This also includes some mechanisms to offer Quality of Service (QoS) guarantees to be widely available at the IP layer. Moreover, employing GMPLS-based control plane has also created a new paradigm in optical-based service offerings to the end-users. From a traditional static-based, it is now possible to provide dynamic, on-demand, and multiple classes of services with discrete level of service guarantees.
IMCOM '18, January 5–7, 2018, Langkawi, Malaysia From protocol perspectives, GMPLS provides a framework where existing IP-based protocols such as OSPF and RSVP [3] are able to function seamlessly in the optical domain. Lightpaths, which are the main entity of the connections that exist within the networks, can be established by utilizing these protocols. From service perspectives, GMPLS can also provide traffic engineering (TE) for lightpaths within the network. TE capable optical network enables the network to focus on its performance with regard to delivering services at a much fine granularity, which in turn results in a better capacity optimization and utilization [4]. In this respect, much finer QoS classifications at the optical level can be achieved to target specific users' requirements. As stated in [5], V Kauffman mentioned that “network service providers are looking beyond today's clunky service delivery model. Its reliance on manual provisioning system and long-term capacity allocation has made the service model an inefficient one. Next generation network architecture is envisioned where its ability to utilize network capacity intelligently, coupled with the capability to deliver more value-added services will undoubtedly shape the service delivery model into a more beneficial one for both users and providers”. In other words, it is imperative for future network service delivery model to be dynamic and flexible.
3 NETWORK MODELING 3.1
Network Model Concept
The network model used in this works has been entirely developed using OMNeT++ (Objective Modular Network Testbed in C++) discrete-event simulation platform [6]. OMNeT++ is an object-oriented modular test-bed simulator, whereby each module in the network is implemented as an object. Additionally, it supports hierarchically nested modules with flexible module parameters. Therefore, OMNET++ provides a suitable platform that supports modeling of distributed mesh topologies [7]. The network topology of interest in this work is meshed based configuration which consists of a set of OXCs, connected by a set of paired fiber links. Dedicated channel in each link consists of the control plane topology in which control messages are exchanged independently from the data plane topology. Internally, an OXC consists of two main parts: a control plane and a data plane. The control plane consists of three units: the signaling, the routing, and the admission control unit. The functionalities of the signaling and routing units are implemented using standard GMPLS protocols as described in the Internet drafts and implemented in [8]. This work focuses on the implementation of the admission control unit functions in conjunction with other protocols. The data plane is essentially a hardware foundation. It is responsible for switching the incoming traffic to an appropriate output wavelength, according to the information presented in the wavelength routing table and wavelength status. Each OXC has a controller model in-charge of controlling and scheduling all required functions in the node. It is also responsible for delivering messages to remote nodes through the dedicated control channel; which are essentially GMPLS protocol standard
2
F. Ali et al. messages generated locally or passed over to other nodes by the signaling unit. OXC units require particular information in order to efficiently implement their functionality. Furthermore, the amount of information significantly impacts on model scalability. Three data tables were used in the OXC implementation. • Wavelength routing table: contains the information that describes the status of wavelengths at each port. • Lightpaths information table: maintains the information about all lightpaths generated from, or terminated on, the corresponding OXC. Specifically, the lightpath route information is maintained here. This information enables the control plane, to change or modify the lightpath route in order to improve network performance. • Network Physical topology table: contains the information about the entire network link connectivity. This information enables the routing unit to calculate the appropriate route for a new request. Lightpath connections are requested and terminated randomly with requests arriving based on a Poisson process. The lightpath parameters include the source, the destination, and the service class, selected randomly based on a uniform distribution. The routing units determine the shortest path based on the number of free wavelengths in each link. The first fit wavelength assignment strategy was considered with wavelength continuity constraint condition is met. Lightpath provisioning employs the parallel reservation scheme using the GMPLS signaling protocol [9]. Based on this scheme, a connection request is forwarded from the source to the destination and the intermediate nodes, and collects the resource information on its way. Based on this information, the source then selects the appropriate label (wavelength) and sends a reservation request to corresponding nodes. The connection request will be blocked if there are no available resources along its route.
3.2
Network Model Structure
The network structure employed in this work is the WDM network architecture. The critical element of the network model is the optical node, as explained in the previous section. Figure 1 illustrates the elements in an optical node figuratively. In this work, the data plane consists of two main topologies namely link and lightpath topologies; while the control plane is made up from the GMPLS control protocols [10]. These are interpreted into the routing and resource discovery unit, signaling unit, and the admission control unit. The controller model which resides in the control plane utilizes a dedicated control channel to manage the operation on each node. This control channel is used to deliver messages to remote nodes using the GMPLS standard protocol messages implemented by the cMessage class under Omnet++ simulation platform. These messages are then passed other nodes by the signaling unit. The approach in this work is to implement an admission control units to manage the lightpath requests at each optical node. This admission control unit has full view of the connections established in the networks, primarily on the connections’ source-destination pairs and the requested classes of services.
Performance of Wavelength Routed Optical Networks Employing GMPLS-based Control Plane
IMCOM '18, January 5–7, 2018, Langkawi, Malaysia assignment procedures are initiated and calculated based on the current network state. Upon completion, the source node generates a number (n) of PATH messages which are sent to all the downstream nodes along the computed route. Nodes receiving the PATH messages then check the wavelength availability selected by the originating node. Subsequently, each particular node generates RESV message and send it to the source node if the wavelength is available. In the event of the unavailability of the selected wavelength, a RESERVE_ERROR message is generated instead and sent to the source node indicating that the lightpath request cannot be accommodated. If the source node receives back exactly n RESV messages, the node controller then configures the OXC with the appropriate wavelength and, consequently, the lightpath is established.
PATH RESV
Figure 1: The internal components of an optical node.
3.3
Routing and Resource Discovery Unit
The integral purpose of this unit focusses on the implementation of the routing and wavelength assignment (RWA) algorithm [11]. The components and entries of the physical topology table and link state table, as well as the wavelength routing table are completed by this unit. Link state information between the OXCs is exchanged and updated using the Open Shortest Path First (OSPF-TE) [4] routing protocol. This unit generates and receives the link state advertisements (LSA) which contain the information about network resources; in terms of the link status and wavelength usage. When a connection is established or deleted, this unit is also responsible for LSA flooding to advertise this fact to the rest of network nodes to update their link state databases. Additionally, the routing element is in-charge of finding the optimal route when a new request arrives. The choice of a path prompted by a connection request is made by taking into account on the current link state information which also includes wavelength availability.
3.4
Control Admission Unit
This unit is created to support the process for admitting connection requests into the network. In future, this unit can also address the implementation of connection provisioning process in optical networks model. The main element in this unit is the global resources information on the network model. Subsequently, it enables each network node to have full awareness on the resource utilization in the network. The overall approach of lightpath setup process implemented in this work is illustrated in Figure 2. When a connection request arrives at the source node, path computation and wavelength
RESV PATH
Figure 2: Connection provisioning process for establishing lightpaths between source-destination pairs.
4
SIMULATION EXPERIEMENTS
Figure 3 (a) and (b) are used to test the network model implementation with different network configurations. In Figure 3(a), 4-node network configuration is used; consisting of string, ring, partial mesh and full mesh topologies. While Figure 3(b) consists of 16-node configuration with ring, star-ring, partial mesh and torus mesh topologies. In the simulation, the following network parameters are assumed:
3
IMCOM '18, January 5–7, 2018, Langkawi, Malaysia
F. Ali et al.
• Connection request arrives into the network according to Poisson process with λ connections per second. • Connection request is in uniform distribution; which means that source-destination pair is determined randomly. • Connection holding time is in exponential distribution with predetermined mean; μ second. • In 4-node configuration, each link consists of a bi-directional fiber with 4 wavelengths per link; and 16 wavelengths per link in the 16 node configuration • Network load is measured in Erlang; denoted as the ratio between the connection inter-arrival rate and the connection holding time. Connection blocking or blocking probability is chosen as the performance metric to evaluate the network model performance. It refers to the probability that a connection cannot be established due to resource non-availability along the desired route; i.e. there is no common wavelength between a source-destination pair. This blocking probability is calculated as a ratio between rejected connection requests and the total requests.
Figure 4: Blocking probability for 4-node networks which consists of string, ring, partial mesh and full mesh topologies.
Ring
Stt ng String g Rin Rin Ring
Star-Ring
tiall Mesh Partial Partial Mesh
Torus Mesh
lll Mesh M Full
(a)
(b)
Figure 3 (a) and (b): 4-node and 16-node network configuration used in the simulation experiments. The system performance derived from the model implementation of both 4-node and 16-node configurations are depicted in Figure 4 and Figure 5, respectively. These results are consistent with the common assumption on the blocking probability in the communication network. Network topologies with higher degree of connectivity provide more available paths to accommodate requests, thus increasing its probability to be successfully established. In contrast, networks with lower degree of connectivity suffer from limited paths, or require longer path lengths to have the connection requests established. As illustrated in the figures, the overall blocking in the network decreases as more path diversities exist in the networks. On both cases, mesh networks give the best performance in terms on blocking ratio for connection request arrivals in the network.
4
Figure 5: Blocking probability for 16-node networks which consists of ring, star-ring, partial mesh and torus mesh topologies.
5
CONCLUSIONS
In a wavelength-routed optical network, GMPLS-based control plane has been employed to automate the connection provisioning throughout network operations. While building a testbed for experimentation is an expensive approach, a simulation and modeling approach is best suited for performing performance analyses for such networks. Using Omnet++ simulation platform, a network model of wavelength-routed optical networks has been developed and tested rigorously. The results obtained from the experiments conducted show that the network model has demonstrated a consistent behavior in terms of its connection blocking probability. This fact can effectively pave ways for more performance analysis works on such network architecture. These prospective works include employing differentiated service classes with respects to the connection requests. Besides connection blocking probability, integrating the OXC’s switching time on the data plane can efficiently indicate the delay and response time for each connection request. These aspects shall be investigated in the future.
Performance of Wavelength Routed Optical Networks Employing GMPLS-based Control Plane
REFERENCES [1] M. Ritter, 2009. Automated Service Delivery for Dynamic Optical Networks, ADVA Optical Networking, August, 2009. [2] A. Farrel, I. Bryskin, 2006. GMPLS: Architecture and Applications, Morgan Kaufmann Publishers. [3] B. Ramamurthy and B. Mukherjee, 1998. Wavelength Conversion in WDM Networking, IEEE Journal on Selected Areas in Communication - Special issue on high-capacity optical transport network, vol. 16, no. 7, September 1998, pp. 1061-73. [4] D. Katz et al., 2003. Traffic Engineering (TE) Extensions to OSPF Version 2, RFC 3630, Sept 2003. [5] V. Kaufmann, 2005. Optical control plane will rewrite the rulebook, Lightwave Magazine (Optical Technologies, Communication Applications, and Industry Analysis Worldwide), November 2005.
IMCOM '18, January 5–7, 2018, Langkawi, Malaysia [6] A. Varga, 2001. The OMNeT++ Discrete Event Simulation System, In the Proceedings of the European Simulation Multiconference (ESM'2001), Prague, Czech Republic, June 6-9, 2001. [7] L. Begg et al., 2006. Survey of Simulators of Next Generation Networks for Studying Service Availability and Resilience, Technical Report TR-COSC 05/06, University of Canterbury, Christchurch, New Zealand. [8] S. Al Barrak et al., 2016. Spare Capacity Modeling and its Application in Survivable IP-over-Optical Networks, European Modeling and Simulation Symposium, Barcelona, Spain. [9] L. Berger, 2003, Generalized Multi-Protocol Label Switching (GMPLS) Signaling RSVP-TE Extensions, RFC 3473, Jan 2003 [10] C. Xin, et al., 2001. On an IP-Centric Optical Control Plane, IEEE Communication Magazine, September, 2001, pp. 88-93. [11] M. Sengupta, S. Mondal, D. Saha, 2012. A Comparison of Wavelength Reservation Protocols for WDM Optical Networks, Journal of Network and Computer Applications, March 2012.
5
