Label swapping versus label preserving in packet switched networks: Impact on the label space size

https://doi.org/10.1016/j.osn.2008.06.001Get rights and content

Abstract

In recent years, switching and networking solutions exploiting all-optical nodes are gaining increasing interest to achieve the target of ultrawide-bandwidth and low-latency packet or burst processing. On the one hand, many prototypes, validated by experimental demonstration of all-optical label processing solutions, have been developed. On the other hand, the primitive available technology for performing label processing poses several constraints on the label structures; this in turn significantly impacts the traffic engineering aspects of such a network.

In this paper, the label assignment problem is studied in a network that makes forwarding decision based on optical packet labels and formulated independently of any technology. Specifically, the problem of assigning labels to identify the label switched path (LSP) packets in a unique and disjoint way is defined, with the objective of optimally minimizing the label space size (i.e. number of labels, or bits, required to uniquely identify the LSPs). The network scenarios where (i) labels have local point-to-point significance (i.e. the label is swapped when traversing each node), and (ii) labels have end-to-end significance (i.e. the label is preserved along the LSP traversing multiple nodes) are both investigated. For both scenarios, labels can be uniquely identified at each node or at each node-port. The label assignment strategies for all the possible scenarios are investigated.

Both theoretical and practical methods, i.e. integer linear programming formulations and heuristics, respectively, are used to assess the efficiency of the proposed label preserving solution. Numerical results show that, for a significant set of network topologies, the label space size increase experienced by networks with label preserving capabilities is limited or negligible in both per-node and per-port label identification.

Introduction

Optical packet switching advances are pushing the technology from the electronic processing domain to the optical processing domain. As a result of this evolution, the capabilities of multiprotocol label switching (MPLS) or generalized MPLS (GMPLS) protocols to perform traffic engineering with quality of service guarantees are going to be provided by means of the all-optical processing of the packet labels.

Optical node architectures that implement optical packet processing have been proposed and experimentally tested [1], [2], [6]. The proposed node architectures have the capabilities of optically processing the packet labels and switching the optical data packets accordingly. In addition, the node architectures may also offer the additional capability of swapping labels in the optical domain, i.e. modifying on-the-fly the optical incoming labels into a possibly different outgoing label [2], [3], [4], [5].

Although the feasibility of label swapping nodes or LS nodes, i.e. optical nodes with label swapping capabilities, is being pursued and evaluated, a number of issues complicate LS node implementation and utilization in the optical domain. The reason is that label swapping functionalities require additional optical components for label generation and insertion [6], [1], making the LS node architecture more expensive [7], more complex [6], and in turn more prone to failure. Moreover, the optical label signal, that undergoes a number of label swaps, can be eventually affected significantly by noise and degradation.

To avoid or limit these adverse effects, [6] proposes a label stripping strategy. The packet header contains a label for each one of the node that the packet must traverse, i.e. it contains a set of node-identifier labels. Label stripping nodes switch the packets according to their outermost labels, that are then stripped off from the header. The thorough study in [6] demonstrates that label stripping strategy requires smaller size labels (i.e. node identifiers) and thus is able to reduce the optical components, required in the proposed implementation of the label stripping nodes. As pointed out in [6], the proposed approach has a main weakness when implemented in a real network. The header length increases with the number of nodes traversed by the packets (the numerical estimation of the overhead is included in [6] in Table 10). This drawback is accentuated in networks with Quality of Service (QoS) requirements. To ensure QoS, each label should also include additional bits, reserved for the class of service field or the differentiated service field. Therefore, the advantages of the small label size may be lost or paid with a higher overhead.

In this paper, a different strategy is proposed to avoid or at least limit the drawbacks of label swapping. The aim is to reduce the optical domain complexity and tasks, by removing the label swapping functionalities from the optical nodes. Under such scenario, the optical nodes, referred to as label preserving nodes or LP nodes, switch data packets all-optically according to their labels, without modifying (e.g. swapping) or removing them. This implies that, in a network of LP nodes, or LP-capable network, the optical data packets are constrained to keep the same label from the source to the destination node, i.e. they undergo the label-continuity constraint.

While avoiding the drawbacks of label stripping strategy, label preserving strategy introduces an increase of the complexity and tasks of the control plane in LP-capable networks. Indeed, not only the control plane has to set up the (optical) label switched paths (LSPs), along which data packets are optically routed, but also it has to appropriately assign an end-to-end label to each LSP and to inform the interested nodes about the selected labels. In particular, label assignment is critical due to the label-continuity constraints and may results in a larger number of labels, required to support end-to-end LSPs, with respect to LS-capable networks. Indeed, the number of optical components required at the node may be a function of the number of labels to be handled by the node. Therefore, LP-capable network solution may become practical and cost-effective, only if the the increase of the number of labels may be kept contained.

In this paper, the label assignment problem in LP-capable networks is formulated, assuming full knowledge of the LSP requests. The problem consists in assigning end-to-end labels to requested LSPs in a unique and disjoint way, such that the label space size, i.e. number of labels (or bits) required to uniquely identify the LSPs, is minimized. Uniqueness and disjointness can be defined at each node or at each node port. Both label assignment strategies are explored.

The objective of the work is to evaluate whether, in LP-capable networks, (1) the additional complexity introduced in the control plane (i.e. label assignment problem) is treatable and (2) the label space size is kept limited compared to a LS-capable networks. The label assignment problem is formally defined for LP-capable and LS-capable networks when nodes are unique per-node and per-port. The complexity of the label assignment problem is evaluated as NP-complete, but sub-optimal solutions can be found in polynomial time by resorting to graph coloring heuristics. The sub-optimal label space size of LP-capable networks is then compared against the label space size of LS-capable networks, for various network topologies and LSP routing scenarios. Results are also compared against header size requirements of label stripping solution. The estimation of the label space size requirements in LP-capable networks helps to evaluate whether LP-capable networks can be a viable solution for the future all-optical packet switched networks, based on connection-oriented paradigm.

Section snippets

Label assignment example

The constraints imposed by the LP and LS nodes on the label assignment are explained in this section with the help of an example.

Fig. 1, Fig. 2 show an LP node. In Fig. 1, two label-continuous LSPs pass through the LP node. Packets of each LSPs carry the same end-to-end label. In this case, labels are identified on a per-node basis, according to the node forwarding table. Therefore, for a proper packet switching, the label assigned to the packets of an LSP should be distinct from the label

Label assignment problem formulation

This section focuses on the label assignment problem and evaluates the minimum number of labels, or label space size, required in LP-capable and LS-capable networks. In the following, a single wavelength is considered and the ports are, thus, fiber links. The problem can be easily extended to consider also the wavelength domain [6].

Given

  • G(V,E): graph representing the network physical topology. V is the set of nodes representing the LS or LP nodes. E is the set of unidirectional links

Label space size comparison

In this section, the label space size in LS and LP-capable networks with per-node and per-port label identification is compared for different physical topologies and LSP routing. The considered physical network topologies, along with their number of nodes, N, and links, L, are indicated in Table 1. Fig. 5, Fig. 6, Fig. 7 graphically represent the considered mesh topologies.

An LSP is requested between each node pair. The LSP routes, forming set P, are selected among the K shortest paths, such

Conclusions and discussion

To reduce the optical node complexity and to avoid the signal degradation and the noise introduced by the optical label removal and insertion, the label preserving strategy is proposed for all-optical label-switched networks. By using label preserving strategy, the optical label swapping functionalities are removed from the nodes, at the expenses of a potential increase of both the network control plane complexity and label space size. This article evaluated the impact of using label preserving

Acknowledgements

The authors would like to express their gratitude to Nicola Calabretta for the the valuable technical input.

References (13)

  • D.W. Matula et al.

    Graph coloring algorithms

  • N. Calabretta, G. Contestabile, A. D’Errico, E. Ciaramella, All-optical label processing techniques for pure DPSK...
  • F. Ramos, et al. IST-LASAGNE: Towards all-optical label swapping employing optical logic gates and optical flip-flops,...
  • D.J. Blumenthal

    All-optical label swapping networks and technologies

    Journal of Lightwave Technology

    (2000)
  • E.N. Lallas, N. Skarmoutsos, D. Syvridis, An optical FSK-based label coding technique for the realization of the...
  • A. Chowdhury, Y. Jianjun, C. Gee-Kung, All-optical label swapping for same wavelength data switching using optical...
There are more references available in the full text version of this article.

Cited by (0)

This work has been supported by MUR under Italy-Tunisia FIRB project “Software and Communication Platforms for High Performance Collaborative GRID (RBIN043TKY)”.

View full text