Dynamic lightpath restoration is a key issue in wavelength switched optical networks (WSON). On the other hand, the introduction of the path computation element (PCE) and the generalized multi-protocol label switching (GMPLS) architectures into WSON as control plane technologies is expected to bring more intelligence and to enable the dynamic computation and control of end-to-end lightpaths in a cost-efficient manner. In this paper, for the first time and through a lab trial with four domains and a field trial located in Japan and Spain, we experimentally present PCE-based optical signal to noise ratio (OSNR)-aware dynamic restoration in multi-domain GMPLS-enabled translucent WSON, assessing the overall feasibility of the proposed techniques and quantitatively evaluating the service disruption time and path computation latency during end-to-end lightpath restoration.
©2011 Optical Society of America
Translucent wavelength switched optical networks (WSON), with sparse 3R regenerators deployed to alleviate the wavelength continuity constraint (WCC) and to guarantee the signal quality of a lightpath, seek a graceful balance between opaque and transparent networks. For an intelligent path computation and dynamic lightpath provisioning in such translucent WSON, two different approaches are usually considered. One is a fully distributed solution based on the generalized multi-protocol label switching (GMPLS) architecture, in which the ingress node for a connection request is responsible for the impairment-aware routing and wavelength assignment (IA-RWA). The other one is to introduce a dedicated path computation element (PCE) to perform IA-RWA. In both approaches, the GMPLS protocol suite is responsible for topology (and resource) dissemination through routing protocol (i.e. open shortest path first-traffic engineering, OSPF-TE) and the actual provisioning of label switched paths (LSPs) through the signaling protocol (i.e. resource reservation protocol-traffic engineering, RSVP-TE).
Lightpath restoration is a key issue in WSON, since a single link failure may result in the loss of a large amount of data. According to , there are two typical restoration schemes in a GMPLS-enabled network. The first one is often referred to as pre-planned restoration, where the backup path is pre-computed by GMPLS (or PCE) and automatically set up after a link failure is detected. The second one is known as dynamic restoration. In this scheme, the restoration path is dynamically computed by GMPLS (or PCE) after the failure of a link, according to the up-to-date network status. To this end, a lot of studies have been carried out in recent years related to both the pre-planned and dynamic restoration both in fully distributed GMPLS controlled WSON or PCE/GMPLS controlled WSON; some of the experimental studies are summarized in Table 1 . It can be seen that, despite massive progress, the dynamic restoration in PCE/GMPLS-controlled multi-domain WSON has not been addressed yet, especially through experimental approaches.
In light of this, in this paper, we summarize the PCE-based intra-/inter-domain path computation algorithms, the GMPLS protocol extensions and the interworking between PCE and GMPLS control plane. Based on these techniques, we experimentally assess the feasibility and efficiency of PCE-based dynamic restoration in GMPLS controlled translucent WSON, through a lab trial with four domains, as well as a field trial located in Saitama (Japan) and Barcelona (Spain).
The rest of this paper is organized as follows. Section 2 presents the technical details. The lab trial setup, the field trial setup and the experimental results and discussions are detailed in Section 3. Section 4 concludes this paper by summarizing our contributions.
2. PCE-based OSNR-aware dynamic restoration in multi-domain GMPLS-enabled translucent WSON
In this section, the PCE-based path computation algorithm is described, and the enabling GMPLS protocol extensions are summarized. Finally, the interworking between PCE and GMPLS, as well as the signaling procedure for PCE-based dynamic restoration are investigated in detail.
2.1 PCE-based OSNR-aware multi-domain path computation
In our implemented PCE, we select the optical signal-to-noise ratio (OSNR) as the signal quality performance indicator. The rationale behind this choice has been investigated in detail in our previous work .
The multi-domain path computation is based on the backwards recursive path computation (BRPC) mechanism . With BRPC, PCEs collaborate to compute the optimal end-to-end path: the destination domain PCE computes a virtual shortest path tree (VSPT) from the domain ingress nodes to the destination node, and then sends the computed VSPT to the upstream PCE in order to compute its own VSPT. Each upstream PCE recursively applies this procedure up to the source domain, obtaining an optimal path from the source node to the destination node.
Within each domain, the deployed path computation algorithm is based on a modified Dijkstra. The basic principle of this algorithm is presented in Fig. 1 and the details of this algorithm are presented in . In short, it computes the path, the regeneration points and the wavelengths at each transparent segment, minimizing the path cost, and fulfilling both the WCC and the OSNR requirements.
2.2 GMPLS protocol extensions
To support the aforementioned PCE-based path computation, and given the fact that the PCE constructs its traffic engineering database (TED) directly from the OSPF-TE packets as detailed below, we extend the OSPF-TE type 10 opaque link state advertisement (LSA) to dynamically disseminate wavelength availability at each link. Moreover, we introduce new link and node type-length-values (TLVs) in OSPF-TE to advertise link and node OSNR (definitions of link/node OSNR are given in ) as well as 3R regenerator information (e.g. availability, tunable range) within each domain. In addition, to set up an end-to-end lightpath based on the PCE path computation results, we extend RSVP-TE to explicitly specify and control the wavelength to select for each strict hop and the nodes wherein the 3R regeneration is to be performed, by means of a label sub-object and a regenerator bit in the explicit route object (ERO), as we studied in [5,8,11].
2.3 Interworking between PCE and GMPLS
The interworking between the PCE and the GMPLS controller is depicted in Fig. 2 . The implemented PCE is capable of dynamically parsing the extended OSPF-TE packets, processing the information in the link and node TLVs and constructing a dedicated TED. For the dynamic interworking, the GMPLS controllers are extended with path computation client (PCC) capabilities. Once a lightpath setup request (with the source and destination node IP addresses) is received by the GMPLS controller, the RSVP-TE module sends a path computation request (PCReq) message to the PCC module. In turn, the PCC performs the requested PCE communications protocol (PCEP) handshake  with the PCE server and forwards the path computation request (PCReq) message to it, getting the path computation reply (PCRep) obtaining the strict ERO after the end-to-end path computation is completed. Finally, the PCC configures the ERO in the GMPLS controller for an automatic lightpath establishment, and proceeds to signal the lightpath, as shown in Fig. 2 and detailed next.
2.4 Signaling procedure for PCE-based dynamic restoration
We illustrate the scenario in Fig. 3 to describe the signaling procedures for PCE-based dynamic restoration. Assume that a working path (1→2→5→7) is calculated by PCE A and PCE B and is successfully established. When a link fails (e.g. link Node2-Node5), the egress transponder TPND2 detects the loss of light (LOL) and then Node7 immediately sends out a NOTIFY message to Node1 (ingress node), triggered by the failure alarm. Upon receiving this NOTIFY message, Node1 releases the working path and requests a restoration path to the PCE, sending a PCReq message with an exclude route object (XRO) containing the resources to exclude. Note that this XRO needs to be sent in the request downstream so the downstream domain PCEs can apply BRPC correctly. Upon completion of the BRPC procedure, PCE A replies with a PCRep message after the end-to-end restoration path has been computed. The GMPLS controller of Node 1 automatically establishes the restoration path according to the ERO encapsulated in the PCE response. Note that, in general, the XRO can either reflect the record route object (RRO) of the working LSP to achieve an end-to-end failed-route-disjoint restoration, or only contain the failed link for a failed-resource-disjoint restoration. Although the latter can achieve better resource utilization, it requires fault localization mechanisms (e.g., deployment of optical performance monitors for every node, introduction of link management protocol, etc.), which may greatly increase the network complexity in a multi-domain scenario, as well as introduce longer restoration latency. Therefore, in this paper, we use the first solution to generate the XRO for an end-to-end dynamic restoration.
3. Experimental setup, results and discussions
In this section, we present the experimental setups, including a lab trial and a field trial. Then, we report and discuss the experimental results.
3.1 Lab trial setup
To evaluate the feasibility and efficiency of PCE-based dynamic restoration, we constructed a translucent WSON multi-domain testbed with four domains as shown in Fig. 4 . Each domain (i.e. autonomous system) was deployed with a dedicated PCE and permanent PCEP adjacencies existed between neighboring PCEs. Domains 1 and 2 were deployed with both a GMPLS-based control plane and a real physical data plane: in domain 1, two wavelength cross-connect (WXC) nodes and two reconfigurable optical add/drop multiplexer (ROADM) nodes based on wavelength selective switches were used, while in domain 2, four photonic cross-connect (PXC) nodes integrated with dense wavelength division multiplexing (DWDM) MUX/DEMUX optical filters were deployed. Two OTU2 based transponders were attached at WXC1 and PXC2 respectively, and a shared 3R regenerator was deployed at WXC2. Domain 3 was deployed with only a GMPLS control plane and the data plane was emulated. Domain 4 was emulated within the PCE, using a static Japan 14-node topology which is detailed in .
For simplicity, four wavelengths were assumed available in each link in all the domains. We also assumed that in domain 3, each boundary node was equipped with a 3R regenerator and, in domain 4, all the nodes were deployed with a 3R regenerator. The assumed link OSNR values are shown in Fig. 4 and node OSNR values were set to 31.9dB for all the nodes. The minimum required OSNR threshold at the receiver to consider a path feasible was set to 19dB.
3.2 Field trial setup
Besides the lab trial as we described in the above sub-section, we also set up a field trial between Saitama (Japan) and Barcelona (Spain) in order to obtain valuable insights when the proposed techniques are being potentially deployed into real operational scenario, and to validate / insure inter-operability. The field trial setup is shown in Fig. 5 . A dedicated PCE was deployed at both KDDI R&D Laboratories and at CTTC premises, located in Saitama (Japan) and Barcelona (Spain) respectively. For simplicity, they will be referred to as PCE-K and PCE-C respectively in the remainder of this paper. The PCE-K obtained a copy of the TED by dynamically reading and parsing OSPF-TE packets, and for PCE-C, we statically configured a network topology (with emulated data plane), since we focused on path computation functions, as shown in Fig. 5. In this field trial, the PCE-K and PCE-C were connected through the public Internet.
3.3 Experimental results and discussions
We firstly verified the feasibility of PCE-based OSNR-aware dynamic restoration between domains 1 and 2. As shown in Fig. 4, an inter-domain working path from TPND1 to TPND2 was firstly established, and after a link failure was introduced between ROADM1 and PXC1, the end-to-end restoration path was automatically calculated by the PCEs and established through GMPLS signaling, following the aforementioned procedures. Note that the link WXC1-WXC2 was configured with low OSNR, therefore the path segment WXC1-ROADM2-WXC2 was selected rather than WXC1-WXC2 in domain 1 as a part of the restoration path. In addition, the regenerator at WXC2 might be used if the potential path could not meet the WCC or OSNR requirements. Figure 6(a) shows the Wireshark capture of the RSVP-TE message sequence. In this procedure, after the GMPLS controller of WXC1 sent the PATH TEAR message to release the working path, it immediately called the PCC for restoration path computation. Figure 6(a) shows that around 109 ms were required for WXC1 to receive the ERO of the restoration path, which included internal processing latency within the GMPLS controller of WXC1, the round-trip-time among the GMPLS controller, PCC and PCE, and the PCE path computation latency. Figure 6(b) shows the received optical power at TPND2 during the lightpath restoration. The service disruption time measured by the egress SDH analyzer was 913.95 ms (Fig. 6(c)) comprising of signaling latency (~898 ms, as shown in Fig. 6(a)) and the configuration delay of transponders and switches (~16 ms).
Clearly, the PCE-based path computation latency is a major performance indicator for the dynamic restoration, given its impact on the end-to-end restoration latency. Therefore, we quantitatively measured the path computation latency for 1000 restoration requests in the multi-domain testbed, as the results are summarized in Table 2 . The first line of Table 2 indicates the sequence of PCE chain for restoration path computation. It can be seen that, with the increase of the network scale and the number of domains, more time was required to compute an end-to-end restoration path. Figure 7(a) shows the Wireshark capture of PCEP sequence for a restoration path computation request, which is exactly the same as the procedure we investigate in Fig. 3. It should be noted that the XRO is required to be inserted in the PCReq message, as shown in Fig. 7(b).
For the field trial, similarly, we measured the restoration path computation latency by randomly generating 1000 requests with random XRO (one every 10 seconds on 2011/09/30). Figure 8(a) shows the Wireshark capture of the PCEP sequence for a restoration path computation request during this field trial. The computation latency measured by the Wireshark was around 306.91ms. Figure 8(b) shows the round trip time between PCE-K and PCE-C, which was approximate 300ms. From these results, it can be observed that in the field trial, the restoration path computation latency is mainly determined by the round trip time between the neighboring PCEs assuming persistent PCEP connections, otherwise the TCP and PCEP handshakes may severely impact the restoration delay. Figure 8(c) shows the distribution of restoration path computation latency for these 1000 requests. The average value for this latency, as we measured, was around 332.57 ms.
Another observation in both the lab trial and field trial was that if multiple requests for restoration path computation arrived at the PCE in a very short period (e.g., simultaneous LSP failures), the PCE may assign same resources for different requests since they shared a same snapshot of the TED, which will degrade the successful ratio of restoration. Although this issue is a nature of the stateless PCE, we believe that some mechanisms (e.g., a more intelligent IA-RWA algorithm) may be beneficial for addressing this issue, and it will be our future work.
In this paper, we have detailed the successful deployment of PCE-based OSNR-aware dynamic restoration in GMPLS-enabled multi-domain translucent WSON, through a lab trial with four domains and a field trial located in Japan and Spain. We have verified the overall feasibility of the proposed techniques and quantitatively evaluated the service disruption time and path computation latency during end-to-end lightpath restoration. We hope the work presented in this paper would be beneficial for the industrial deployment of WSON with an intelligent PCE/GMPLS-based control plane in the future.
The authors of KDDI R&D Labs would like to thank Dr. Hideaki Tanaka, Dr. Masatoshi Suzuki, and Dr. Shigeyuki Akiba for their continued encouragement. A part of work of CTTC was funded by the MICINN (Spanish Ministry of Science and Innovation) through the Cecyt project DORADO (TEC2009-07995), and by the European Community’s Seventh Framework Programme (FP7/2007-2013) through the Integrated Project (IP) STRONGEST under grant agreement no 247674.
References and links
1. J. P. Lang, Y. Rekhter, and D. Papadimitriou, eds., “RSVP-TE extensions in support of end-to-end generalized multi-protocol label switching (GMPLS) recovery,” IETF RFC 4872 (2007), http://tools.ietf.org/html/rfc4872.
2. I. Nishioka, L. Liu, S. Yoshida, S. Huang, R. Hayashi, K. Kubo, and T. Tsuritani, “Experimental demonstrations of dynamic wavelength path control and highly resilient recovery in heterogeneous optical WDM networks,” in Proceedings of 15th OptoElectronics and Communications Conference (OECC 2010), paper PD1 (Institute of Electrical and Electronics Engineers, New York, 2010), 1–2.
3. S. Huang, L. Liu, S. Yoshida, I. Nishioka, R. Hayashi, K. Kubo, and T. Tsuritani, “Multi-failure restoration demonstrations with multi-vendor interoperability in control plane enabled WSON,” in Optical Fiber Communication Conference and Exposition and National Fiber Optic Engineers Conference (OFC/NFOEC 2011), Technical Digest (CD) (Optical Society of America, 2011), paper NThC6, http://www.opticsinfobase.org/abstract.cfm?URI=NFOEC-2011-NThC6.
4. L. Liu, T. Tsuritani, R. Casellas, and M. Tsurusawa, “Demonstration of a resilient PCE/GMPLS controlled translucent optical network,” in Proceedings of 15th OptoElectronics and Communications Conference (OECC 2010), paper 7A1–2 (Institute of Electrical and Electronics Engineers, New York, 2010), 88–89.
5. L. Liu, T. Tsuritani, R. Casellas, R. Martínez, R. Muñoz, M. Tsurusawa, and I. Morita, “Experimental assessment of a resilient PCE/GMPLS controlled translucent wavelength switched optical network,” IEICE Trans. Commun. E94-B, 1831–1844 (2011).
6. R. Muñoz, R. Casellas, and R. Martínez, “Experimental evaluation of dynamic PCE-based path restoration with centralized and distributed wavelength assignment in GMPLS-enabled transparent WSON networks,” in Proceedings of 36th European Conference and Exhibition on Optical Communication (ECOC 2010), paper Mo.2.D.4 (Institute of Electrical and Electronics Engineers, New York, 2010), 1–3.
7. H. Guo, T. Tsuritani, S. Okamoto, and T. Otani, “Demonstration of GMPLS-controlled inter-domain transparent optical networks,” in Proceedings of 34th European Conference and Exhibition on Optical Communication (ECOC 2008), paper Th.1.E.2 (Institute of Electrical and Electronics Engineers, New York, 2010), 1–2.
8. L. Liu, T. Tsuritani, R. Casellas, R. Martínez, R. Muñoz, and M. Tsurusawa, “Experimental demonstration and comparison of distributed and centralized multi-domain resilient translucent WSON,” in Proceedings of 36th European Conference and Exhibition on Optical Communication (ECOC 2010), paper We.7.D.3 (Institute of Electrical and Electronics Engineers, New York, 2010), 1–3.
9. R. Martínez, R. Casellas, L. Velasco, F. Agraz, R. Muñoz, and S. Spadaro, “Experimental validation of end-to-end GMPLS-enabled restoration in multi-domain transparent WSON,” 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 NThA5, http://www.opticsinfobase.org/abstract.cfm?URI=NFOEC-2010-NThA5.
10. L. Velasco, F. Agraz, R. Martínez, R. Casellas, S. Spadaro, R. Muñoz, and G. Junyent, “GMPLS-based multidomain restoration: analysis, strategies, policies and experimental assessment,” J. Opt. Commun. Netw. 2(7), 427–441 (2010). [CrossRef]
11. R. Martínez, R. Casellas, R. Muñoz, and T. Tsuritani, “Experimental translucent-oriented routing for dynamic lightpath provisioning in GMPLS-enabled wavelength switched optical networks,” J. Lightwave Technol. 28(8), 1241–1255 (2010). [CrossRef]
12. R. Casellas, R. Martínez, R. Muñoz, and S. Gunreben, “Enhanced backwards recursive path computation for multi-area wavelength switched optical networks under wavelength continuity constraint,” J. Opt. Commun. Netw. 1(2), A180–A193 (2009). [CrossRef]
13. J. P. Vasseur and J. L. Le Roux, eds., “Path computation element (PCE) communication protocol (PCEP),” IETF RFC 5440 (2009), http://tools.ietf.org/html/rfc5440.