A constrained-lower-indexed-block spectrum assignment policy in elastic optical networks

https://doi.org/10.1016/j.osn.2019.03.001Get rights and content

Abstract

Elastic Optical Networks have been the most promising technology in the next generation of transport networks due to their flexibility of spectrum assignment. Routing and spectrum assignment is one of their key issues. Several routing and spectrum assignment algorithms have been introduced in recent years. In this research, we propose a novel spectrum assignment algorithm called constrained-lower-indexed-block policy. This policy is based on K shortest paths and has the capability to assign a lower indexed spectrum block on some a path to a connection request by the constraint that, length of the path should not increase too much. Simulation results show that the proposed policy can greatly reduce the blocking probability compared with two well-performed policies.

Introduction

With the rapid development of internet, various emerging bandwidth-hungry applications are making the demand of bandwidth not only to be larger and larger but also more and more diverse. The traditional WDM optical networks can provide large enough bandwidth, but they cannot cope up with the diversity of bandwidth demand as they can only provide channels with fixed wavelengths. A new optical network architecture named SLICE [1] or elastic optical networks (EONs) [2] has been designed and refined in recent years, which can satisfy not only the bandwidth requirement but also the diverse requirements of applications.

In EONs, channels are based on flexible grids rather than fixed wavelengths, so spectrum can be elastically assigned to applications according to their traffic rates. In general, spectrum in EONs is divided into slices (or frequency slots) with equal bandwidth (bandwidth of a slice is much narrower than that of a wavelength in WDM networks) [3], and an application can use several adjacent slices (unlike in WDM networks only one wavelength).

EONs have become a focus in recent years and routing and spectrum assignment (RSA) is one of their important issues. In the process of assigning spectrum slices to a connection request, two constraints need to be considered [4]:

  • Spectrum continuity along a light-path. For a connection request, the spectrum slices on all links of the source-destination path must be completely the same.

  • Spectrum contiguity on each link. A connection request may use several spectrum slices and these slices must be contiguous on each link of the source-destination path.

The former is similar to the wavelength continuity constraint for routing and wavelength assignment (RWA) problem in WDM all-optical networks, while the latter is an additional constraint in EONs. The extra constraint makes the RSA problem in EONs more complex than the RWA problem in WDM networks.

In terms of whether different modulation formats can be used, routing and spectrum assignment problems can be classified into two categories [5,6]: ordinary routing and spectrum assignment problem (still denoted as RSA) and routing, modulation and spectrum assignment problem (denoted as RMSA) [7,8]. In RSA problem, modulation format is fixed and data rate of each slice is constant. In RMSA problem, modulation levels can be adjusted; a shorter path can use a higher modulation level causing a higher data rate. However, the RMSA problem has a higher requirement of transceivers at nodes, which leads to a higher capital expenditure of network deployment. In this paper, we mainly focus on the RSA problem.

Several algorithms have been designed for the RSA problem. Most of them are based on shortest path or K-shortest paths routing algorithms and their main differences are reflected in the rules of spectrum assignment. In this paper, we propose a spectrum assignment algorithm named Constrained Lower Indexed Block (CLIB) policy for the RSA problem. The policy is based on K shortest paths routing algorithm and its main idea is to assign slices with lower indexes on some path to a connection request, with the constraint that the length of path should not be increased too much. The CLIB policy can get an obvious improvement in blocking probability.

The remainder of this paper is organized as follows. In Section 2, related works are summarized. In Section 3 the proposed spectrum assignment algorithm is introduced in detail. Simulations are done in Section 4 to evaluate the performance of CLIB policy. And finally a brief conclusion is drawn in Section 5.

Section snippets

Related works

According to different application scenarios, RSA problems can be classified into two types: online RSA problem and off-line RSA problem.

Off-line RSA problem is in the provisioning stage of EON. In which, a traffic matrix among nodes is given in advance, and we need to arrange the traffics by using as fewer slices as possible. The off-line RSA problem can be converted to an integer linear programming (ILP) model and the objective is to minimize number of required slices [9].

Online RSA problem

Proposed algorithm

By comparing the spectrum assignment policies, we find that if a policy can let occupied slices be as concentrative as possible, it would get a good performance of blocking probability. However, the existing methods either fail to consider this idea or only consider this idea on one path, thus the effects are limited. In this paper, we design a spectrum assignment algorithm which extends the idea to a wider range.

The proposed spectrum assignment algorithm is based on K shortest paths algorithm.

Performance evaluation

We investigate performance of the proposed policy in this section. We first examine the impacts of various parameters to blocking probability of the CLIB policy and then compare performance of the CLIB policy with that of two well-performed policies. Simulation settings are listed in Table 3.

Conclusions

In this paper, we propose a spectrum assignment policy named constrained lower indexed block (CLIB) for the RSA problem in EONs. The policy is based on K shortest paths, and its core intention is to reduce spectrum fragments and then decrease blocking probability. To achieve this effect, for two blocks on different paths, the policy would choose the lower indexed block on the longer path rather than the higher indexed block on the shorter path. This measure would aggregate busy slices at lower

References (26)

  • R. Zhu et al.

    Multi-path fragmentation-aware advance reservation provisioning in elastic optical networks

  • Y. Wang et al.

    A study of the routing and spectrum allocation in spectrum-sliced elastic optical path networks

    Proc. - IEEE INFOCOM

    (2011)
  • B. Mukherjee

    Optical WDM Networks[M]

    (2006)
  • Cited by (26)

    • A prime-partition strategy for spectrum assignment problem in elastic optical networks

      2021, Optical Fiber Technology
      Citation Excerpt :

      Other constraint-unaware algorithms are specifically designed for the RSA problem in EONs. For example, the first-last-fit algorithm divides the frequency slices into several sections and the first-fit rule is used in odd sections while the last-fit rule is used in even sections [25]; the constrained-lowest-index-first algorithm extends the first-fit rule to K-shortest paths case to get a lower index frequency block [26]; the tradeoff-RnI algorithm extends the first-last-fit rule to K-shortest paths case and makes a tradeoff between distance to boundary and consumed resources [27]. The constraint-unaware algorithms adopt some kinds of heuristic ideas to optimize the use of spectrum resources.

    View all citing articles on Scopus

    This work is supported by Natural Science Foundation of China No.61501406 and 61772477, Higher Educational Scientific Research Program of Henan Province No.15A510015 and 15A520032.

    View full text