Applying pcycle protection for a reliable IPTV service in IPoverDWDM networks
 Ahmed Frikha^{1}Email author,
 Bernard Cousin^{1} and
 Samer Lahoud^{1}
DOI: 10.1186/1869023853
© Frikha et al.; licensee Springer. 2014
Received: 8 April 2013
Accepted: 28 January 2014
Published: 4 March 2014
Abstract
Today, television over Internet protocol (IPTV) has become very popular and service providers must deal with the rapid growth in the number of IPTV customers. Service providers must ensure the reliability of IPTV to satisfy customers’ needs, as a network failure could disrupt an IPTV transmission.
Survivable multicast routing is important for providing a reliable IPTV service. Generally, most carriers route multicast traffic using the protocolindependent multicast sourcespecific mode (PIMSSM) based on the routing information provided by the interior gateway protocol (IGP). Restoration using the IGP reconfiguration is slow and typically takes from 10 to 60 seconds. To ensure a fast restoration, we consider node and link failure recovery in the Dense Wavelength Division Multiplexing (DWDM) optical layer. The backup path is provided in this layer. Thus, the multicast tree does not change at the IP layer (the logical links do not change) and the restoration time is faster (typically of the order of 50 to 80 ms).
In this paper, we apply pcycles in IPoverDWDM networks to provide a robust IPTV service. In addition, we propose a novel concept for node protection using pcycles to achieve more efficient resource utilization. We also propose a new algorithm, the node and link protecting candidate pcycle based algorithm (NPCC). This algorithm integrates our new concept for node protection. Extensive simulations show that it outperforms the existing approaches in terms of blocking probability, resource utilization efficiency and computation time rapidity.
Keywords
IPTV service multicast routing IPoverDWDM networks Reliability pcycles; node and link protection1 Introduction
Nowadays, many telecoms companies offer the television over IP (IPTV) service and distribute television channels using backbone networks. An IPTV service requires stringent quality of service constraints (for example, for packet loss, jitter and endtoend delay) to satisfy customers’ needs. Service providers must also ensure the reliability of the IPTV service. A simple link or router failure could disrupt the television content distribution for several seconds, if no protection mechanism is implemented.
IPTV contents could be carried using the IP multicast to save bandwidth capacity. In fact, multicasting enables a single packet to be sent to multiple destinations at once. Although many multicast routing algorithms have been proposed for IPTV services, most of the carriers today implement the protocolindependent multicast sourcespecific mode (PIMSSM) [1]. For many reasons, this protocol is efficient for internet broadcaststyle applications such as IPTV. Obviously, PIMSSM is simple to implement for network operators thanks to the sourcespecific mode (SSM) [1], which makes this protocol ideal for IPTV. SSM does not require the network to maintain knowledge about which sources are actively sending multicast traffic, unlike the internet standard multicast (ISM) protocol. This advantage makes PIMSSM more scalable than ISM.
To ensure IPTV reliability, survivable multicast routing must be guaranteed. Moreover, service providers must ensure a fast restoration time for link and router failure recovery. The PIMSSM protocol uses the routing information provided by the interior gateway protocol (IGP) to compute a multicast tree. Thus, restoration at the IP layer using IGP reconfiguration requires IGP to be aware of the failure, then PIMSSM can use the new IGP shortest paths to rebuild a new multicast tree using the pruneandjoin process. This operation is slow, and typically takes from 10 to 60 seconds [2]. To avoid rebuilding the multicast tree and to ensure a fast restoration, we consider node and link failure recovery at the DWDM layer. The backup path is provided at this layer. This makes restoration time faster as the multicast tree does not change at the IP layer (the logical links do not change).
The pcycle technique was extended to support node protection in the DWDM layer using the node encircling pcycle concept (NEPC) [5]. According to this concept, a protecting pcycle of a given node must link all neighboring nodes of the failed node. This constraint is too strict. The method could discard some nodes that can be protected by the pcycle but which do not satisfy the constraint. This will affect the efficiency of the pcycles in terms of capacity saving.
In this paper, we consider link and node failure recovery at the DWDM layer using pcycles. We extend the node protection concept of the pcycle approach to achieve more efficient resource utilization. Then, we propose a novel algorithm, the node and link protecting candidate pcycle based algorithm (NPCC). The NPCC algorithm integrates our proposed concept for node protection. This algorithm ensures node and link failure recovery based on a set of candidate pcycles to overcome the high computation time problem.
The rest of this paper is organized as follows. In Section 2, we present the IPTV architecture and discuss the restoration mechanisms to ensure a reliable IPTV service. In Section 3, we extend the pcycle protection concept for protecting nodes in light trees. In Section 4, we present our novel algorithm for combined node and link failure recovery using the novel node protection concept. Extensive simulations and numerical results are presented in Section 5. The conclusions are given in Section 6.
2 IPTV architecture and restoration mechanisms
In this section, we present the main components of the IPTV architecture and we give an example. Then, we discuss the restoration mechanisms, and we highlight the advantages of applying the pcycle protection approach for IPTV.
2.1 IPTV architecture
The main components of an IPTV architecture are [2–6]:

A super headend (SHE): The SHE is located in the core network. Also called the IPTV backbone, it collects television content from television networks, such as satellites and offair distributions. After video processing, encoding and management, the SHE distributes the television content using IP routers to multiple video hub offices (VHOs).

A video hub office (VHO): A VHO receives IPTV content transmitted by the SHE through the IPTV backbone routers. Then, it combines this content with the local television and the video on demand (VoD) content. The SHE routers, the VHO routers and the links that connect them form the IPTV backbone. Each VHO in turn serves a metro area by transmitting the IPTV content to multiple video serving offices (VSOs).

A video serving office (VSO): A VSO contains the aggregation routers that aggregate local loop traffic from subscriber homes, i.e., local digital subscriber line access multiplexers (DSLAMs).
PIMSSM is largely used for IPTV video distribution with an IPTV backbone. A multicast tree is computed using this protocol based on the routing information provided by the IGP. The multicast tree is used to deliver IPTV content from the SHE to each VHO. Each television channel is assigned to a unique multicast group.
An IPTV service requires a high bandwidth and stringent quality of service constraints. In particular, IPTV is very sensitive to packet loss, as one lost packet could disrupt video quality. Delay and the jitter are also critical for the quality of an IPTV service. These three quality of service constraints rely on network dependability. A link or node failure could disrupt the IPTV service, if there is no restoration mechanism for the multicast tree.
2.2 Restoration mechanisms
With IGP reconfiguration, after a link or node failure, PIMSSM must rebuild the multicast tree. But before that, IGP must be aware of the failure and compute the new shortest paths at the level of each router. PIMSSM will use these new shortest paths to rebuild the multicast tree using the pruneandjoin process. This approach is not suitable for a realtime IPTV service as the restoration process takes too much time, typically between 10 and 60 seconds [2]. With multiprotocol label switching (MPLS) fast reroute protection, the restoration time can be reduced to between 50 ms and 100 ms [2]. Backup labelswitched paths (LSPs) are preestablished, and stored in the router forwarding tables and this makes fast rerouting possible at the IP MPLS layer.
Other restoration mechanisms have been used in the DWDM layer and can achieve lower restoration times. Oneplusone (1 + 1) and onebyone (1:1) restoration can be implemented in the DWDM layer. The restoration time for these approaches is less than 20 ms [2]. However, they are not efficient in terms of bandwidth saving, as backup paths cannot be shared. In these approaches, a backup path is dedicated for one and only one working path. The pcycle protection approach, described in the previous section, ensures node and link failure recovery while maintaining a fast restoration time (typically of the order of 50 to 80 ms) [4]. Moreover, this approach achieves an efficient use of the network capacity compared to the other protection approaches.
These restoration mechanisms are proposed for unicast traffic. Some other restoration mechanisms focus on protecting multicast trees against network failures. In 2009, F Zhang and WD Zhong proposed the efficiencyscore based heuristic algorithm of node and link protecting pcycle (ESHN) [7]. Although the ESHN algorithm has a lower blocking probability than the OPPSDP algorithm [8] and the ESHT algorithm [9] in dynamic multicast traffic, ESHN does not efficiently use the protection capacity provided by a pcycle, especially when protecting nodes. The ESHN algorithm does not take into consideration all nodes that a pcycle can protect when selecting a protecting pcycle. This is due to the two hard constraints imposed by the method used by ESHN for protecting nodes. The first constraint is that a node protecting a pcycle has to link all onelevel downstream nodes of the failed node. The second constraint is that the pcycle must contain one of the upstream nodes of the failed node in the light tree. Of course these reduce the computation time for the algorithm as they limit the search space for the pcycles. However, they prevent the ESHN algorithm from achieving the best resource utilization. Furthermore, when the traffic load is high, the computation time for the ESHN algorithm remains high and does not satisfy the IPTV service requirements.
In this work, we focus on the design of a reliable IPTV service. We use link and node failure recovery at the DWDM layer to give a short restoration time. We use pcycles to ensure efficient use of network capacity. We also extend the node protection of the pcycle approach to achieve more efficient resource utilization. In Section 3, we provide a detailed study of node protection using pcycles and we present our proposed concept for protecting nodes in the light tree.
3 Node protection using pcycles
3.1 Existing approaches for node protection using pcycles
3.2 The proposed concept for node protection using pcycles
In this section, we present our novel concept for protecting nodes in multicast traffic. Before presenting our concept, we will introduce some notation. Let T be a multicast light tree to be protected, s be the source node in T, N_{ f } be an intermediate node in T, and D={d_{1},d_{2},..,d_{ i }} be the set of destinations in T that are affected when a failure occurs on the node N_{ f }.
Theorem 1
 1.
The node N _{ a }∈T, the node N _{ e }∈T and N _{ f }∉ [ s,N _{ a }] where [ s,N _{ a }] is the segment in T linking the source s to the node N _{ a }.
 2.
∀d _{ k }∈D, ∃ a node N _{ k }∈ [ N _{ a },N _{ e }] and N _{ k }∈]N _{ f },d _{ k }], where ]N _{ f },d _{ k }] is the segment in T linking N _{ f } to d _{ k }.
 3.
N _{ f }∉ [ N _{ a },N _{ e }].
Proof Once a failure occurs on the node N_{ f }, the multicast traffic is rerouted through the pcycle C_{ j } to ensure the survivability of the multicast session. The pcycle must provide a protection segment to deliver the multicast content to all destinations that are affected by the failure of N_{ f }. This segment is denoted by [ N_{ a },N_{ e }].
First, we justify why constraint (1) is required. The extremities N_{ a } and N_{ e } of this segment must be in T. In fact, the node N_{ a } is responsible for injecting the multicast traffic into the protection segment [ N_{ a },N_{ e }] when N_{ f } fails. In addition, N_{ a } must not be affected by the failure of N_{ f }, i.e., N_{ a } must continue to receive the multicast traffic even if a failure occurs on node N_{ f } (N_{ f }∉ [ s,N_{ a }]). The node N_{ a } must split the incoming light signal into two outgoing signals. The first is injected into the protection segment and the second is forwarded to the downstream node of N_{ a } in the light tree T. The node N_{ e } is the last intersection node between T and C_{ j }.
Second, we prove the necessity of constraint (2). To ensure failure recovery, we must make sure that all destinations affected by the failure of N_{ f } continue to receive the multicast traffic through the protection segment [N_{ a },N_{ e }]. Two scenarios are possible for delivering the multicast traffic to an affected destination d_{ k }. In the first, the segment [ N_{ a },N_{ e }] carries the multicast traffic directly to d_{ k }, i.e., the protection segment traverses the node d_{ k }. In the second scenario, the segment [ N_{ a },N_{ e }] carries the traffic to the destination through an intermediate node N_{ k }. The node N_{ k } must be an upstream node of d_{ k } and a downstream node of N_{ f } in the light tree. This constraint ensures that the failed node N_{ f } does not belong to the segment [ N_{ k },d_{ k }] of the light tree. The node N_{ k } splits the incoming signal into two signals. The first is sent to the next node in the protection segment to ensure that the remaining affected destinations will receive the multicast content. The second is forwarded to the downstream node of N_{ k } in the light tree to reach d_{ k }.
Finally, we prove that constraint (3) is necessary. We must make sure that the protection segment is not affected by the failure of N_{ f }. Therefore, the protection segment [N_{ a },N_{ e }] should not traverse the node N_{ f }.
3.3 Example
4 The proposed protection algorithm
In this section, we present our proposed algorithm NPCC for protecting nodes and links in the DWDM layer to give a reliable IPTV service. Our algorithm deploys the aforementioned concept for node protection using pcycles.
4.1 Selection of candidate pcycles
First, the NPCC algorithm enumerates a set of candidate pcycles in an offline phase, i.e., before any requests have been received. Using these candidate pcycles will considerably reduce the computation time for the algorithm. In fact, considering the total pcycle set when selecting a new pcycle to be established, is a very slow task, especially when the number of pcycles in the network is high. Therefore, we select a set of candidate pcycles to reduce the computation time for our algorithm.
A pcycle with a high PC is useful as it maximizes the amount of protected capacity while reserving less spare capacity. The l pcycles with highest PC are selected as the candidate pcycle set, where l is a parameter for the algorithm. The goal in selecting this set is to maximize the capacity that can be protected on the network.
4.2 Flow chart for the NPCC algorithm
The ESbased unity pcycle procedure calculates the ES of each unity pcycle in the candidate pcycle set and selects the pcycle with maximum ES. The set of links protected by the selected unity pcycle is removed from L and the set of protected nodes is removed from N. This process is iterated until all the links and all the nodes of T are protected, i.e. L=ϕ and N=ϕ. The selected unity pcycles are configured and the corresponding wavelengths are reserved. Note that the reserved pcycles may serve to protect subsequent multicast requests. This is why after routing a multicast tree, we compute the set of links in L and the set of nodes in N that can be protected by the existing pcycles in the network. Finally, the reserved capacity of an existing pcycle in the network is released when the pcycle does not protect any working link and nodes in the network.
5 Performance evaluation
In this section, we present the evaluation of our algorithm NPCC, which is proposed for providing a reliable IPTV service. Our method guarantees the recovery from link and node failure at the DWDM layer with a fast restoration time. We compare our algorithm with the ESHN algorithm, which was reported to be the most efficient algorithm for dynamic multicast traffic protection in terms of resource utilization efficiency and blocking probability.
In our simulation, we assumed that request arrival follows a Poisson process with an average arrival rate λ, and the request holding time follows an exponential distribution with an average holding time μ. Hence, the offered traffic load for the network is given by λ μ.
We ran simulations on the following wellknown and frequently used European optical topologies developed within the COST266 [11] and COST239 [12] projects:

The COST266 core topology [11] contains 16 nodes and 23 links, with an average nodal degree of 2.88. The total number of pcycles in this topology is 236 (118 pcycles in each direction).

The COST239 topology [12] contains 11 nodes and 26 links, with an average nodal degree of 4.727. The total number of pcycles in this topology is 5,058 (2,029 pcycles in each direction).
In our study, without loss of generality, we assumed that each link has two fibers. The two fibers transmit in opposite directions; 16 wavelengths are available on each fiber. The source and the destinations of each multicast session are randomly selected (uniform distribution law). We chose the number of destinations in each multicast request D=5, which seems to be reasonable as the total number of nodes in the used topologies is less than 16. We compared the performance of the algorithms using the following performance criteria:

The blocking probability (BP), which is the percentage of requests that cannot be routed or protected among the total number of requests.

The resource utilization (RU), which is the percentage of reserved wavelengths in the network among the total number of wavelength links.$\mathit{\text{RU}}=\frac{{W}_{R}}{E\times W}$
where W_{ R } is the total number of wavelength links reserved in the network, E is the number of fibers in the network and W the number of wavelengths per fiber.

The average computation time (CT), which is required for routing and protecting a traffic request.
Performance criteria BP, RU and CT were computed according to the traffic load. For each traffic load value, 5×10^{5} requests were generated. This number of requests is enough to measure BP, RU and CT with a 95% confidence interval.
6 Conclusion
In this work, we focused on the reliability of an IPTV service. First, we presented the main components of the IPTV architecture, then we discussed existing restoration mechanisms for the IP and DWDM layers. The restoration methods proposed for DWDM are more efficient and more suitable for IPTV in terms of restoration time. We also highlighted the advantage of applying pcycle protection for producing a reliable IPTV service.
Second, we extended the concept of node protection using pcycles to deal with multicast traffic. Our novel concept allows the protection capacity provided by a pcycle to be used efficiently. We proposed a novel algorithm, NPCC, which uses our concept for node protection. The NPCC algorithm ensures both link and node failure recovery in the DWDM layer for dynamic multicast traffic. This algorithm reduces the computation time for setting up a multicast traffic request by enumerating a set of candidate pcycles based on the PC score.
Finally, we compared our proposed algorithm with the ESHN algorithm, which was reported to be the most efficient algorithm for node and link failure recovery for dynamic optical multicast traffic. Extensive simulations showed that the NPCC algorithm had a lower blocking probability and outperformed the ESHN algorithm in terms of resource utilization and computation time.
Abbreviations
 BP:

blocking probability
 CT:

computation time
 DSLAM:

digital subscriber line access multiplexer
 DWMD:

dense wavelength division multiplexing
 ERH:

efficiencyratiobased unity pcycle heuristic algorithm
 ES:

efficiency score
 IGP:

interior gateway protocol
 IP:

Internet protocol
 IPTV:

television over IP
 ISM:

internet standard multicast
 LSP:

labelswitched path
 MPLS:

multiprotocol label switching
 NEPC:

node encircling pcycle concept
 NPCC:

node and link protecting candidate pcycle based algorithm
 PIMSSM:

protocolindependent multicast sourcespecific mode
 QoS:

quality of service
 RU:

resource utilization
 SHE:

super headend
 SSM:

sourcespecific mode
 VHO:

video hub office
 VoD:

video on demand
 VSO:

video serving office.
Declarations
Authors’ Affiliations
References
 Bhattacharyya S: An overview of sourcespecific multicast (SSM). IETF RFC 2003, 3569.Google Scholar
 Doverspike R, Ramakrishnan KK, Chase C: Structural overview of ISP networks. In Kalmanek C, Misra S, Yang R (eds) Guide to reliable internet services and applications, Computer Communications and Networks. London: Springer; 2010:19–93.View ArticleGoogle Scholar
 Grover WD, Stamatelakis D: Cycleoriented distributed preconfiguration: ringlike speed with meshlike capacity for selfplanning network restoration. Proceedings of IEEE International Conference on Communications (IEEE ICC), vol 1 1998, 537–543.Google Scholar
 Clouqueur M, Grover WD: Availability analysis and enhanced availability design in pcyclebased networks. Photonic Netw Commun 2005, 10(1):55–71. 10.1007/s111070051695xView ArticleGoogle Scholar
 Doucette J, Giese PA, Grover WD: Combined node and span protection strategies with nodeencircling pcycles. In Proceedings of the workshop on design of reliable communication networks (DRCN). Italy: Ischia (Naples); 2005:213–221.Google Scholar
 Cisco Systems: White paper on optimizing video transport in your IP triple play network. 2006.http://www.cisco.com .Google Scholar
 Zhang F, Zhong WD: Performance evaluation of optical multicast protection approaches for combined node and link failure recovery. J Lightw Technol 2009, 27(18):4017–4025.View ArticleGoogle Scholar
 Singhal NK, Sahasrabuddhe LH, Mukherjee B: Provisioning of survivable multicast sessions against single link failures in optical WDM mesh networks. J Lightw Technol 2003, 21(11):2587–2594. 10.1109/JLT.2003.819550View ArticleGoogle Scholar
 Zhang F, Zhong WD: p cycle based tree protection of optical multicast traffic for combined link and node failure recovery in WDM mesh networks. IEEE Commun Lett 2009, 13(1):40–42.View ArticleGoogle Scholar
 Zhang ZR, Zhong WD, Mukherjee B: A heuristic method for design of survivable WDM networks with pcycles. IEEE Commun Lett 2004, 8: 467–469. 10.1109/LCOMM.2004.832772View ArticleGoogle Scholar
 Maesschalck SD, Colle D, Lievens I, Pickavet M, Demeester P, Mauz C, Jaeger M, Inkret R, Derkacz BM: PanEuropean optical transport networks: an availability based comparison. Photonic Netw Commun 2003, 5(3):203–226. 10.1023/A:1023088418684View ArticleGoogle Scholar
 Batchelor P, Daino B, Heinzmann P, Weinert C, Spath J, Van Caenegem B, Hjelme DR, Inkret R, Jager HA, Joindot M, Kuchar A, Le Coquil E, Leuthold P, De Marchis G, Matera F, Mikac B, Nolting HP, Tillerot F, Wauters N: Ultra high capacity optical transmission networks. Final report of Action COST 1999, 239.Google Scholar
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License(http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.