We present a Resource Broker with a Virtual Network Topology Manager (VNTM) which dynamically deploys virtual GMPLS-controlled WSON networks. Virtual Optical links are constructed by grouping established optical connections which are managed by the VNTM. We evaluate the performance of the Resource Broker in the ADRENALINE testbed.
© 2012 OSA
Optical Network Virtualization involves the dynamic provisioning of dedicated virtual networks over the same physical optical infrastructure, which is attracting a lot of attention from network infrastructure providers, with the purpose of offering their Infrastructure as a Service (IaaS). Optical network virtualization technologies allow the partitioning/aggregation of the network infrastructure into independent virtual resources, where each virtual resource has the same functionality as the physical resource . Virtual Optical Networks (VON) support the heterogeneous and stringent infrastructure network requirements of the emerging dynamic and bandwidth-hungry applications such as high-definition video streaming and cloud computing. Thus, service providers can dynamically request, on a per need basis, a dedicated VON for each application and have full control over it. In order to independently provide the functionalities of automatical optical connection provisioning, traffic engineering and dynamic protection/restoration to the resulting virtual instances, a VON must be composed of not only a virtual transport plane but also of a virtual control plane.
The authors have previously presented a VON Resource Broker architecture for deploying GMPLS-controlled VONs , with the limitation that the requested virtual optical links needed to be directly mapped into a physical optical link (a virtual optical link can span more than one physical optical link). To overcome this limitation and establish the requested virtual optical links, the authors in  have proposed a Resource Broker which includes a Virtual Network Topology Manager (VNTM). The proposed Resource Broker with a VNTM dynamically deploys virtual GMPLS-controlled WSON networks. This VNTM communicates with a Control Plane consisting of GMPLS and Path Computation Element (PCE) in order to construct the requested virtual optical links as a set of LSPs and the VNTM will offer the established LSPs as virtual TE links for the virtual GMPLS control plane.
This paper extends the presented results in , by introducing Global Concurrent Optimization (GCO), which allows the PCE to perform concurrent path computations, where a set of paths are computed concurrently in order to efficiently utilize the network resources. GCO allows the concurrent path computation of all the requested virtual optical links in order to perform an efficient VON Resource Allocation (VON RA). We provide a GCO algorithm in order to evaluate the performance of a concurrent path computation for the requested virtual optical links. The paper also introduces the VON Resource Broker with VNTM system architecture and experimentally assesses and evaluates the proposed system architecture on the ADRENALINE testbed, by providing performance results (i.e., blocking rate of VON requests and VON setup delay).
2. System architecture
Figure 1 shows the proposed system architecture. A virtualizable GMPLS/PCE-controlled WSON network is managed by the proposed Resource Broker, which is the responsible for managing the incoming asynchronous and dynamic VON requests, which consists on the allocation of the new VON, the modification of resources assigned to an existing VON or the releasing of the resources in case a VON is torn down.
A VON request is modeled as a graph that describes a set of virtual optical switches and links for the virtual transport plane, specifying for each one the number of requested input and output optical ports, and the number of wavelengths, respectively. The VON request also includes some requirements and constraints for the virtual control plane, such as the needed capacities for the virtual GMPLS controllers, or the selected values for configuring the parameters/attributes of the control processes running on the virtual GMPLS controllers, which can be later modified by the service provider. Internally, the Resource Broker consists of 5 modules, namely, VON Controller, Resource Manager, Resource Allocator, VNT Manager and Resource Configurator.
The Resource Manager handles the virtual control resources. It manages the available IP subnetworks that shall be used to establish the virtual IP Control Channel (IPCC) to later deploy dedicated Data Communication Network (DCN) for each virtual transport plane. It is also responsible for managing the number and location of the available virtual GMPLS controllers that each physical GMPLS controller supports (static partitioning), as well as their configuration information (including the management IP address, the amount of CPU power and the available RAM). It also stores all the information required to configure the processes (e.g., routing, signaling, etc.) running in the virtual GMPLS controllers.
The VON Controller accepts incoming TCP sessions, used to reliably transport VON requests, and handles these requests asynchronously and dynamically. Once the VON identifier is assigned, or found, the VON controller triggers the resource allocator in order to process the VON request.
The VNTM is composed of a Path Computation Client (PCC) and a LSP Manager. Using the GCO, we group all the requested virtual optical links into a Path Computation Request for the PCE. Each requested virtual optical link includes the requested number of wavelengths. The PCE will reply with a Path Computation Reply, which includes Explicit Route Objects (ERO). A virtual optical link can be defined in one or more EROs (depending on the requested number of wavelengths). Each ERO is used by the Connection Controller to request the necessary LSPs through the GMPLS connection controller (i.e., RSVP-TE protocol) that is running on the source node of each requested virtual link. Once each LSP has been established, the RSVP-TE protocol answers with a Record Route Object (RRO). The established virtual optical link is offered as a virtual TE link to the virtual GMPLS control plane, which is configured to map the virtual TE link to the physical port of the optical node (i.e., ROADM or OXC).
The Resource Allocator assigns the control resources to the requested VON. For the virtual control plane, it allocates the virtual GMPLS controllers, and assigns the GMPLS router address. It also assigns IP addresses and GRE tunnels for the required IPCC.
The Resource Configurator generates the virtual transport and control plane configuration XML file, which describes a VON scenario model that can be set up, modified or torn down by means of ADRENALINE Network Configurator (ADNETCONF) . This is a proprietary software tool in charge of scenario model management in ADRENALINE testbed . Using ADNETCONF, the scenario model is then serialized to the formal representation of the scenario that the processing engine understands. Up to five different XML files are produced; one describing the logical DCN topology for the virtual control plane, and the others describing the configuration of the different GMPLS processes.
3. Experimental assessment and performance evaluation
In this section we present a experimental assessment and a performance evaluation of the proposed system architecture on the virtualizable GMPLS-controlled WSON platform of the ADRENALINE Testbed .
3.1. Experimental assessment
The GMPLS-controlled WSON platform of the ADRENALINE Testbed is composed of an all-optical WSON infrastructure with 2 ROADMs and 2 OXCs providing reconfigurable (in space and in frequency) end-to-end lightpaths, deploying a total of 610 km of G.652 and G.655 optical fiber, with six DWDM wavelengths per optical link. Each optical node is equipped with a virtualization server running in a Linux-based router with an Intel Core 2 Duo E6550 2.33 GHz processor.
Figure 2 shows protocol details on the message exchange for VON request and provisioning, including the PCE request and reply for VON resource assignment, VNTM XML protocol for LSP request at the GMPLS-controlled nodes and LSP establishment (RSVP-TE) messages. As an example, we propose a VON request consisting of 2 virtual optical links between nodes 1 and 2 (i.e., 1–2 and 2-1) (see Fig. 2, right). The links are considered unidirectional, for simplicity. With the purpose of obtaining a path for the requested virtual optical links, the Resource Broker issues to the PCE a Path Computation Request (PCRequest) message including an SVEC object, where the different requested links are enumerated (Fig. 2, left). In this example, two END-POINT objects are requested, including source and destination. As an example, the PCE uses the Shortest Path Unreserved Bandwidth First-Fit (SPUB-FF) algorithm, which consists on computing a constrained shortest path algorithm for each requested virtual optical link. SPUB-FF only takes into account the unreserved bandwidth. Once the spatial paths for the virtual optical links have been computed, the common set of available wavelengths of the computed paths are obtained, in order to overcome the Wavelength Continuity Constraint (WCC). The WCC requires that the same set of assigned wavelengths is allocated on all of the requested virtual optical links. The First-Fit (FF) wavelengths of the common set of available wavelengths of the computed paths are assigned. The PCE replies with a Path Computation Reply (PCReply) message which includes the EROs (spatial path) and the possible wavelengths (i.e., labelset), for each virtual optical link.
The PCReply message is processed and, through the LSP manager, an LSP is requested through an XML proprietary interface to the RSVP process of the source node (Fig. 2, center). Once a setup LSP request message is received at node 1, a RSVP PATH message is issued to node 2, following RSVP standard procedure. Node 2 responds with RSVP RESV message which includes the RRO, and when the message is received at node 1, the optical resources for the LSP have been occupied. Finally, node 1 sends to the Resource Broker an acknowledgement for the setup of the LSP. The LSP establishment is performed for all the requested virtual optical links and the requested wavelengths.
Once all the necessary LSPs have been established, the necessary virtual GMPLS control plane resources need to be allocated. To this end, the established LSPs are used as virtual TE links. Finally, once the VON configuration is generated, the VON Configurator, by means of ADNETCONF, is the responsible to set up or tear down the requested VON. We observe that the virtual GMPLS-controlled WSON setup and tear down delay for our testbed are 17s and 7s, respectively.
3.2. Experimental performance evaluation
We have implemented and evaluated the proposed SPUB-FF algorithm in the the GMPLS-controlled WSON platform of the ADRENALINE Testbed. The inter-arrival process is Poisson, and the holding time (HT) follows a negative exponential distribution. The average Inter-Arrival Time (IAT) is set to 5s. and the average VON HT is varied for an offered traffic load ranging from 1 to 400 Er. Each VON requests 1 wavelength. 103 VONs have been requested for each data point. The VON request topology is generated randomly selecting the number of nodes (V ∈ [2–4]) and the number of links (E ∈ [1–6]). To grant that a connected graph is generated we follow the next procedure, starting with one node. Then, we iterate, creating a new node and a new link. The link is to connect the new node with a random node from the previous node set. As only bidirectional VONs make sense, we also add a link between the random node to the new node. After all nodes are created, we create random links until the VON request is fulfilled.
Figure 3 depicts the blocking rate of VON requests in the NSFNet topology scenario for the proposed VON RA SPUB-FF algorithm. For a given VON request load of 1 Er., the obtained blocking rate of VON requests is 18.2%. This higher bloking rate is due to the fact that several of the randomly constructed requests cannot satisfy the WCC. For a VON request load of 5 Er. the VON blocking rate is of 26.3%.
We define the VON setup delay as the time required to allocate (i.e., does not include establishment time) a virtual WSON. Figure 4 shows that under a VON request load of 1 Er. the VON setup delay is of 104.7 ms. The VON setup delay decreases with higher VON request load. This is due to the fact that with higher VON request loads, the PCE allocates virtual optical links with a lower number of spatial path hops, decreasing the VON setup delay.
We have presented a Resource Broker with VNTM which acts as interface between service providers and infrastructure providers to deploy virtual GMPLS-controlled WSON infrastructure services. Virtual optical links are established as the grouping of established optical connections on the virtualizable GMPLS-controlled WSON and they are offered as virtual links to the virtual GMPLS control plane. Experimental evaluation carried out in the ADRENALINE testbed has shown the feasibility of deploying independent instances of virtual GMPLS-controlled WSON, providing low delays for VON setup and tear down.
We have presented the SPUB-FF algorithm and evaluated its performance in terms of blocking rate of VON requests and VON setup delay. We propose further study on more complex VON RA algorithms, taking into account GCO, which will lead to a more efficient usage of the available physical optical network resources.
This work was supported partially by the Spanish Ministery of Economy and Competitivity through the project DORADO (TEC2009-07995).
References and links
1. R. Nejabati, E. Escalona, S. Peng, and D. Simeonidou, “Optical network virtualization,” in Proc. of Optical Network Design and Modeling (ONDM), 2011, http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5753389.
2. R. Vilalta, R. Muñoz, R. Casellas, and R. Martinez, “Experimental demonstration of a virtual optical network resource broker and compositor for dynamic GMPLS WSON infrastructure services,” in Proc. of OFC/NFOEC, 2012, http://www.opticsinfobase.org/abstract.cfm?URI=OFC-2012-OM3G.4.
3. F. Galán and R. Muñoz, “An automatic model-based reconfiguration and monitoring mechanism for flexible GMPLS-based optical networking testbeds,” in ONDM Lecture Notes in Computer Science 4534/2007, 239–248, http://www.springerlink.com/content/9v0m633420853021/.
4. R. Muñoz, C. Pinart, R. Martinez, J. Sorribes, M. Maier, A. Amrani, and G. Junyent, “The adrenaline testbed: integrating GMPLS, XML, and SNMP in transparent DWDM networks,” IEEE Commun. Mag 43(8), 40–48 (2005). [CrossRef]
5. R. Muñoz, R. Casellas, R. Martinez, R. Vilalta, J. Vilchez, and J. Vazquez, “Virtualizing adrenaline testbed for deploying dynamic GMPLS-controlled wson as a service,” in Proc. of OFC/NFOEC, 2011, http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5875311.
6. R. Vilalta, R. Muñoz, R. Casellas, and R. Martinez, “Dynamic virtual GMPLS-controlled WSON using a resource broker with a VNT Manager on the adrenaline testbed,” in Proc. of European Conference on Optical Communications (ECOC), 2012.
7. Y. Lee, JL. Le Roux, D. King, and E. Oki, “Path Computation Element Communication Protocol (PCEP) requirements and protocol extensions in support of global concurrent optimization,” IETF RFC5557, 2009, http://tools.ietf.org/html/rfc5557.