Abstract-New networking protocols are desired essentially in future Optical Wide Area Network (OWAN) environments to meet increasing demands of communication with low latency and high bandwidth. In this paper, a simulation study based on 100 nodes random topology is conducted to observe packet delay patterns and efficient link utilization for two session traffic sources Exponential and CBR (Constant Bit Rate), with number of wavelengths (16, 32, 48 and 64) while varying packet sizes ranging from 100 to 1000 bytes. By using OWns (Optical WDM network simulator), an optimal packet size is presented for Exponential session traffic as one of the design considerations of new protocol in Wavelength Routed Optical Networks (WRONs) with Wavelength Division Multiplexing (WDM) technology. Simulation results also show that Exponential session traffic performs better as compare to CBR session traffic in terms of average packet delay and link utilization.
Index Terms-Average Packet Delay, Link Utilization, Optical WDM network simulator (OWns), Wavelength Routed Optical Networks (WRONs)
---------- ïµ ----------
1. Introduction
FIBER communications have already a dominant role as a truly broadband networking transport infrastructure due to their ability to grow gracefully, especially after the emergence of the WDM technology [5]. Optical WDM networking technology has been identified as a suitable candidate for future WAN environments due to its potential ability to meet rising demands of low latency and high bandwidth communication [1]. Optical networks are high-capacity telecommunications networks based on optical technologies and components that provide routing, grooming and restoration at the wavelength level as well as wavelength-based services. WDM or optical transport networks have assumed important significance due to their ability to carry large amounts of data traffic. In such networks, data is transmitted from its source to its destination in optical form. Switching and routing operations are solely performed in the optical domain without under doing any optical-to-electrical conversion. In the absence of any wavelength conversion a connection (also called a lightpath) is supposed to use the same wavelength on all links along the chosen path [4].
By using OWns, this simulation study is conducted to investigate effects of varying packet size on average packet delay and link utilization with Exponential and CBR traffic sources in WRONs. Further simulations are performed to find effects of different wavelength numbers on these two network performance parameters; namely, average packet delay and link utilization. This information can be useful while designing new networking protocol(s) to meet the upcoming operational requirements in future optical WANs with WDM technology.
----------------
* Asima Nisar is Assistant Professor in Dept. of Computer Science & I.T., Federal Urdu University of Arts, Sciences and Technology, Gulshan Campus, Karachi - 75300, Pakistan. E-mail: [email protected]
Briefings of layered architecture of optical WDM networks, topology of Optical WDM networks, and overview of OWns including its architecture, components and functionalities are discussed in section 2 as background. In section 3, related work is discussed. Further, simulation study with scenario configuration, simulation parameter values and OWnam's visual output is presented here as well. Simulation results and relevant graphs are discussed in section 4. Conclusion and future work are mentioned in section 5. References are listed in section 6.
2. Background
2.1 Optical WDM Networks: Layered Architecture
The WDM network consists of atleast an IP layer and an optical layer. According to ITU-T (International Telecommunications Union - Telecommunication standard), deployment of WDM technology introduces a new layer in the layered hierarchy, named 'Optical Layer'. Optical layer consists of a set of lightpath established on the physical topology of the network. Optical layer provides circuit-switched lightpath service to higher its layers, such as SONET/SDH, ATM and IP [3].
Optical Channel layer provides wavelength routing and switching functions to establish end-to-end optical connections between network nodes for transparent data delivery. Optical Multiplex layer is responsible for the aggregation of multiple signals. Optical Transmission Layer handles transmission of optical signals on different types of optical fibers such as single-mode and/or multi-mode fibers [3].
Fig. 1. Layered network [3]
2.2 Topology of Optical WDM Networks
The topology of WDM networks is an important architectural issue that needs careful attention. The mesh configuration is more flexible than the other options. Cost reduction in 10 Gb/s optics would make the optical mesh architecture even more attractive. So, the next generation Internet backbone would be a flexible, reconfigurable and reliable mesh optical network with OXCs (Optical Cross Connects) connected to one another [3]. Optical mesh networks will enable a variety of dynamic services such as bandwidth-on-demand, just-in-time bandwidth, bandwidth scheduling, bandwidth brokering, and optical virtual private networks that open up new opportunities for service providers and their customers alike [4].
Fig. 2. Wavelength Routed WDM Mesh Network [3]
2.3 OWns: Overview
OWns (an extension of network simulator ns2; designed especially to simulate optical WDM networks) has been developed as a generic framework to study routing in Optical WDM networks. It provides a definite support for modeling of WDM network characteristics, generating random network topology with (Exponential, CBR, and Pareto) traffic patterns, evaluating performance of new protocols, and making possible to visualize simulation events/sub-events with an option of extensibility in all of its features. For this OWns has two scenario generation tools, namely topology generator and traffic generator. Topology generator creates random topologies according to a set of specified parameters with a definite number of wavelengths. Traffic generator generates random traffic source destination pairs, according to specified traffic models and parameters [1].
OWnam, an extension to nam of ns, is designed to visualize WDM network scenarios for circuit-driven environment. Mainly, it has two components, one is events monitor and other is virtual topology statistics. Event monitor is used to capture and display dynamic events that occur in virtual topologies. Virtual topology statistics is used to visualize the dynamic information of virtual topology construction involving the current state of virtual topology, the state of established lightpaths, and wavelength usage on multi-wavelength links [1].
2.3.1 OWns: Architecture and Layers
OWns architecture encompasses the key characteristics of WDM networks including optical switching nodes, multi-wavelength links, and RWA (Routing and Wavelength Assignment) algorithms. OWns adopts a certain level of abstraction to build the specific switching schemes of WDM networks (e.g. circuit switching) based on the packet switching framework of ns2. A new class of traffic sources termed the session traffic is implemented to generate traffic sessions suitable for WDM circuit switching simulations. The traffic generator randomizes source and destination pairs according to their uniform distribution. By default, all generated multi-wavelength links have the same wavelength number and all wavelengths have the same bandwidth.
Fig. 3. OWns architecture and layers [1]
OWns view the physical and logical topology of WDM networks being implemented as the physical layer and the logical layer respectively. It uses C++ to implement efficient building blocks (such as nodes, links, traffic models, and existing protocol suites) as well as transmission mechanisms and a scripting language OTcl, an object-oriented extension of Tcl as a glue (simulation description languages that configure simulation scenarios). The current version of OWns supports circuit switching [1].
2.3.2 OWns: Components and Functionalities
OWns circuit-switched architecture is composed of routing module, WA module, optical switching nodes, and the multi-wavelength links. Optical switching node, multi-wavelength link, routing module, and WA (Wavelength Assignment) module are implemented as WDMNode, duplex-FiberLink, RouteLogic/ Wavelength, and WAssignLogic objects respectively. Multi-channel structures of multi-wavelength links are centrally maintained in the 'Logical Layer'. WA module works along with routing module to compute wavelength assignment, set up lightpaths, and construct the virtual topology. Relying on these results, optical nodes forward incoming traffic to the corresponding next hops through multi-wavelength links [1].
3. Related Work and Simulation Study
In order to meet bandwidth demand and decreasing cost, many researches have been done in the domain of switching modes in optical network [7].
3.1 Optical Switching Techniques
With the tremendous growth of Internet Protocol (IP) traffic and the rapid development of WDM technology, optical fiber switching techniques have been promoted over recent years. Up to now, there are mainly three kinds of optical network switching techniques being raised, namely, Optical Circuit Switching (OCS), Optical Packet Switching (OPS), Optical Burst Switching (OBS).
3.1.1 Optical Circuit Switching (OCS)
OCS is perfectly fit for large, stable and long duration traffic flows where the lightpath setup overhead is amortized over a large amount of traffic and thus can be ignored. So, OCS can easily support QoS (Quality of Service) and traffic engineering. In OCS networks, lightpath for a large flow requires dedicated wavelengths on all links. In this way, it cannot perform multiplexing of wavelength with higher bandwidth utilization [6].
3.1.2 Optical Packet Switching (OPS)
In OPS, packets are buffered and routed in the optical domain. OPS networks have the switching granularity on the packets level. The functionality of OPS node should include; decoding packet header (can be electronic if the packet header is encoded at lower bit rates), configuring a switch fabric (the reconfiguration needs to be performed very fast in nanosecond range), synchronization (for synchronous OPS nodes), multiplexing and contention resolution. The lack of flexible optical buffers makes the contention resolution in optical domain very difficult [9].
3.1.3 Optical Burst Switching (OBS)
In contrast to OCS, OBS is based on statistical multiplexing, which can increase the efficiency of network resource utilization. OBS networks mainly consist of two types of switching nodes, namely edge and core nodes. The main functions of the edge nodes are optical burst assembly/disassembly, offset time and burst size decision. The OBS core nodes perform control header lookup, optical crossconnecting and data burst monitoring [9].
3.2 Comparison among Optical Switching Techniques
In these modes, OCS is the easiest to deploy, but not efficient to handle burst data (Non-Real-time traffic). Whereas, OPS is the optimal choice, but the necessary optical technologies have not matured yet e.g. optical buffer and optical logic element [7]. Among them, OBS is an integrated one which takes some merits of OCS and OPS while avoiding their demerits [6]. Despite the benefits of OBS paradigm, its high burst blocking probability has delayed its introduction in the industry [8].
3.3 Simulation Study: Why OWns?
As the advent of WDM, hundreds wavelength channels carrying data at rates in order of terabits per second can be multiplexed into a single fiber [2]. Networking protocols and algorithms are being developed to meet the changing operational requirements in future OWANs. Simulation is used in the study and evaluation of new protocols, and is considered a critical component of protocol design. But, a lack of uniformity in the choice of simulation platforms for optical WDM networks makes it difficult for researchers to exchange and compare obtained results under a common simulation environment. To address this need, a WDM network simulation tool called OWns is developed. By using OWns for 100 nodes randomly generated topology, effects of varying wavelength conversion factor are studied on network performance parameters with traffic load 0.3 Erlangs. Similarly, effects of varying traffic load are observed on network performance parameters having wavelength conversion factor 0.5 with the same simulation configuration, mentioned in figure 3 of [1]. In WDM networks, channels are created by dividing the bandwidth into a number of wavelength or frequency bands, each of which can be accessed by the end users at the peak rate their network interfaces can support. In order to efficiently utilize this bandwidth, efficient transport architectures and protocols are needed [5]. This simulation study is conducted to investigate effects of varying packet size with Exponential and CBR traffic sources on average packet delay and link utilization in OWANs with WDM technology. In ns2 [10], Exponential traffic objects generate On/Off traffic. During "on" periods, packets are generated at a constant burst rate. During "off" periods, no traffic is generated. CBR objects generate packets at a constant bit rate.
Further, finding effects of varying number of wavelengths on network performance parameters (average packet delay and link utilization) are targeted also to allow studies on network performance evaluation regarding one of the networking protocol design concerns.
3.4 Simulation Scenario Configuration by using OWns
The default scenario generation tool of OWns is utilized to generate random topology and traffic. Exponential session traffic and CBR (Constant Bit Rate) session traffic sources are used for a 100 node random topology separately under same simulation configurations while varying packet size ranging from 100 to 1000 bytes. The script presented in the simulator as Demo Topology is used to invoke topology generation tool and configure this simulation. Through this script the RWA algorithm with fixed-alternate shortest path routing and first-fit wavelength assignment is evaluated. The wavelength conversion factor, wavelength conversion distance, connectivity probability, traffic pair density and load per traffic pair are set as 0.5, 4, 0.03, 0.6, and 0.3 Erlangs respectively. Wavelength routing is performed on the shortest path.
TABLE 1
SIMULATION PARAMETERS
Topology
Random
Seed to generate Random topology
98765
Wavelength Routing Protocol
WDM Static
Wavelength Assignment Protocol
First Fit
Total no. of nodes
100
Nodes Connection Probability
0.03
Link Wavelength number
16, 32, 48, 64
Link Bandwidth
16 Mb
Link Delay
10ms
Wavelength Conversion Factor
0.5
Wavelength Conversion Distance
4
Wavelength Conversion Time
0.024
Link Utilization Sample Interval
0.5
Traffic Density
0.6
Total Session-Traffics in network
10
Session Traffic Load
0.3 Erlangs
Traffic Arrival Rate
0.5
Traffic Holding Time
1.0
Packet Size
100 - 1000 bytes
Session Traffic Packet Arrival Rate
1 Mb
Traffic Type
Exponential, CBR
Expoo Traffic Average Burst Time
0.7
Expoo Traffic Average Idle Time
0.1
Max. Traffic Requests Number
1000
3.5 OWnam: Visual Tool of OWns
Following is the reflection of visual output of OWnam for the simulation under study.
Fig. 4. Snapshot of the simulation run
In this snapshot shown in Fig. 4, for instance at 461.350770s, a lightpath is created for traffic session 7 from node 5 to node 52 and lightpath is established on the shortest path (path 1) between source and destination without wavelength conversion.
4. Simulation Results and Graphs
100 nodes randomly generated topology with each link having 16, 32, 48 and 64 wavelengths for two session traffic types Exponential and CBR is simulated while varying packet size ranging from 100 to 1000 bytes. Two major network performance measures are strictly observed: average packet delay and link utilization.
4.1 Packet Size vs. Average Packet Delay and Link Utilization for Exponential Session Traffic
Fig. 5a shows the effects of varying packet size on average packet delay with 16, 32, 48 and 64 wavelength numbers along the shortest path for Exponential session traffic.
Fig. 5a
Fig. 5b shows the effects of varying packet size on link utilization with 16, 32, 48 and 64 wavelength numbers along the shortest path for Exponential session traffic.
Fig. 5b
It is observed that for Exponential session traffic, at packet size 300 the average packet delay gets its minimum value with efficient link utilization for each number of wavelengths (16, 32, 48 and 64) in general and particularly 16 wavelength numbers are found as optimal [Fig. 5a], [Fig. 5b].
4.2 Packet Size vs. Average Packet Delay and Link Utilization for CBR Session Traffic
Fig. 6a shows the effects of varying packet size on average packet delay with 16, 32, 48 and 64 wavelength numbers along the shortest path for CBR session traffic.
Fig. 6a
Fig. 6b shows the effects of varying packet size on link utilization with 16, 32, 48 and 64 wavelength numbers along the shortest path for CBR session traffic.
Fig. 6b
It is observed for CBR session traffic, there is a straightforward increase in average packet delay with the increase in packet size while link utilization remains constant showing no effect of varying packet size for each wavelength numbers (16, 32, 48 and 64). Further, 16 wavelength numbers are observed with minimum average packet delay and efficient link utilization [Fig. 6a], [Fig. 6b].
4.3 Number of Wavelengths vs. Average Packet Delay for Exponential and CBR Session Traffic
Fig. 7a shows the effects of varying wavelength numbers on average packet delay along the shortest path for Exponential and CBR session traffics.
Fig. 7a
4.4 Number of Wavelengths vs. Link Utilization for Exponential and CBR Session Traffic
Fig. 7b shows the effect of varying wavelength numbers on link utilization along the shortest path for Exponential and CBR session traffics.
Fig. 7b
5. Conclusion and Future Work
OWns provides the sound platform for evaluating routing and wavelength assignment algorithms and simulating new protocol design concerns in OWAN environment by generating random WDM network topology with 100 nodes.
Packet delay patterns with optimum link utilization are observed and it is presented by using OWns that for Exponential session traffic, at packet size 300 the average packet delay is minimized with efficient link utilization at 16 wavelength numbers.
For CBR session traffic, there is a straightforward increase in average packet delay with the increase in packet size while link utilization remains constant showing no effect of varying packet size at each wavelength numbers (16, 32, 48 and 64). Further, Exponential session traffic performs well while comparing it with CBR session traffic in terms of two network performance metrics; average packet delay and link utilization. This information can be useful while designing new networking protocol(s) to meet the upcoming operational requirements in future optical WANs with WDM technology.
The same simulation study can be repeated with Pareto session traffic for circuit switching networks by using OWns. Further, more realistic traffic sources for WANs with packet switching such as self-similar traffic source with LRD (Long Range Dependence) will definitely be integrated in OWns to achieve valuable simulation results.
