Strategies for the efficient management of the capacity in fixed-mobile converged networks

Nowadays customers can access telecom services via fixed-line networks or via mobile networks. Fixed-line broadband networks in Europe are currently dominated by ADSL technologies which provide up to 16 Mbit/s. Other solutions like Fiber to the Curb (FTTC) with VDSL are widely used with access speeds up to 50 Mbit/s. The technologies which will deliver fiber access directly to the home, referred to as Fiber to the Home (FTTH) solutions, are the next step and first deployments have been started. If the implementation of FTTH is based on a Passive Optical Network (PON) instead of a point-to-point solution, it is possible to reduce the network cost by eliminating the power supply through the installation of passive network elements, and by sharing a significant portion of the network cost among multiple users, while still enabling data rates of hundreds of Mbit/s. For the mobile area, different network technologies are available and widely used. 2G (e.g., GSM, GPRS) and 3G (e.g., UMTS, HSPA) networks have been already installed and in some countries the deployment of the 4G LTE technology (with data rates up to 100 Mbit/s) has been started. Future step is the LTE advanced (LTE-A) technology which can provide data rates up to 1 Gbit/s. With regard to wireless technologies, systems like Wi-Fi (standard IEEE 802.11) or WiMax (standard IEEE 802.16) are also widely used. So far fixed and mobile access networks have been optimised and evolved independently, with some contradicting trends. In fact there is a tendency to centralize fixed networks and to decentralize mobile networks. A certain degree of convergence among the two network domains, typically referred as Fixed Mobile Convergence (FMC), has only been achieved at the service level with the introduction of all IP services (e.g., a practical case of FMC at service level can be found in smartphones and tablets which can access the same services through Wi-Fi and/or the cellular network). It is widely agreed that the development of a single convergent infrastructure for fixed and mobile networks would enable relevant savings in terms of Capex and Opex and would provide converged services to customers at reasonable costs in the years. The development of this FMC access network is driven by the requirement to combine optimal seamless quality of experience for end-users together with an optimised network infrastructure. Moreover, another motivation to merge together in a single and optimized structure both fixed and mobile traffic is related to the energy consumption of the current access network. In fact the access is the part of the network which is consuming the highest amount of energy. The convergence of fixed and mobile access networks in a single structure can help saving energy. In this sense, we can classify fixed mobile convergence referring to two aspects: functional convergence and the structural convergence. The functional convergence refers to the convergence of fixed and mobile network functions, while the structural convergence is the convergence of fixed and mobile infrastructures and equipment. The integration of functionalities and equipment are expected to enable relevant energy saving, e.g, lowering the number of nodes of the access network. One of the most promising network solutions to develop such FMC access networks are Next-Generation Passive Optical Network (NG-PONs). In particular, Long-Reach PONs which use both time-domain and wavelength-domain multiplexing, referred as LR WDM/TDM PON, are a suitable candidate to support a large number of different services (with different QoS requirements) originating from both fixed and mobile users. The aim of this work is to propose, define and technically assess methods to efficiently manage the convergence of Fixed and Mobile traffics over the same infrastructure. In particular, we first proposed Dynamic Bandwidth and Wavelength Allocation (DBWA) mechanisms for Long-Reach WDM/TDM Passive Optical Networks (LR WDM/TDM PONs). These proposed methods have shown to have a better performance, in terms of average packet delay, compared with other existent DBWAs designed for LR WDM/TDM PONs. Moreover, in order to fairly allocate the bandwidth among all the users which transmit over multiple wavelengths in LR WDM/TDM PONs, we proposed a new equation. Such equation, given the length of the polling cycle time (in seconds), provides the maximum amount of bytes that each user can transmit in each polling cycle, according to the capacity assigned to each user. Then, we also studied the effect over the average packet delay of changing the length of the polling cycle. Regarding LR WDM/TDM PONs, we evaluated the multiplexing gain introduced when different number of wavelengths are used. We compared the performance of the different scenarios (with different number of wavelengths) in terms of average packet delay. Such results are also compared with the case where a single wavelength is used to transmit (LR TDM PON). From these results we can provide some preliminary consideration regarding the design of LR WDM/TDM PON which can be used for the FMC backhauling. Then, we studied Dynamic Bandwidth and Wavelength Allocation (DBWA) mechanisms for LR WDM/TDM PONs used in a network scenario where the transmission technologies at the ONUs are all tunable lasers (i.e., uniform transmission technologies scenario). We started by evaluating DBWAs in this uniform transmission technologies scenario to consider which is the impact of the tuning time (TT) of the tunable lasers on the network performance in terms of delay. For this reason, we modified existing DBWAs in order to make them able to consider the TT of Tunable lasers. This modification consists in adding a delay to the transmissions allocated to give enough time to the ONUs to retune their lasers. Then, we evaluated the performance of these modified DBWAs. After this preliminary evaluation, we proposed a new TT-aware DBWA (called TAWA). The fact that a DBWA is TT-aware means that the scheduling is performed considering that the laser needs a certain amount of time to retune to another wavelength (i.e., as a $TT > 0$). Then, since the aim of this work is to study FMC network solutions to efficiently manage the different types of traffic present in such types of networks, we also studied DBWAs for LR WDM/TDM PONs in scenarios where different transmission technologies with different characteristics (i.e., different TTs) are used at the ONUs, considering that different services might be supported by different technologies (e.g., Tunable lasers, array of fixed tuned lasers, fixed tuned lasers, ...). Also in this case, we first evaluated the performance of some existing DBWAs, that we modified to make them considering that different transmission technologies have different TTs, i.e., the modified DBWAS have different allocation policies are adopted for transmissions originated from ONUs with different technologies. In particular, in our work we considered that, in a LR WDM/TDM PON, three types of transmission technologies can coexist: i) Single fixed tuned lasers which cannot retune; ii) Array of fixed tuned lasers which can immediately tune their laser to another wavelength (TT = 0); iii) Tunable lasers which are characterized by a given TT (TT > 0). Then, to understand the importance of having a DBWA which is aware of the exact values of TT and is able to add different delays to different transmissions, we designed a novel version of the existing and already evaluated DBWAs where the OLT does not know the exact value of the TT of each ONUs, but it only knows if a particular transceiver is able to retune or not. These new versions of DBWAs can only exploit a simplified control information, therefore they only know if a transceivers is able to retune or not, but they do not know which is the value of TT that each transceiver needs to retune. For this reason, these new versions of DBWAs add a delay to all the transmissions of the ONUs which have to retune (i.e., Array of fixed tuned lasers and tunable lasers) considering a maximum value of TT (the same for all the ONUs). In this first part of the work, we assumed that the arrays of fixed tuned lasers are able to tune to every wavelength of the network. The number of wavelengths over which such arrays can transmit corresponds to the number of fixed tuned lasers installed in the array. However, we know that to have a cheap device the number of wavelengths where an array of fixed tuned lasers can transmit (i.e., number of lasers of the array) has to be limited. For this reason, we evaluated the performance of the previously studied DBWAs when the arrays of fixed tuned lasers can transmit over a limited number of wavelengths. The results of this study show that using an array of lasers with a limited number of lasers provide an average packet delay which is only slightly higher (in the order of tenth of microseconds) than in the case where the arrays of tunable lasers have full tunability (i.e., they can transmit over every wavelength of the network). Moreover, we proposed a new DBWA that takes into account the fact that real tunable lasers have different values of TT depending on the wavelength they have to retune to (i.e., higher distance between wavelengths requires higher TTs). To the best of our knowledge this is the first time that such a DBWA is proposed and analyzed. Another part of the work concerns more the architecture of the FMC networks. Indeed, in this part of the research we considered a ring topology of an access optical network and we designed an ILP which is able to allocate the resources (i.e., wavelengths, or portion of wavelengths) to the ONUs according to their required amount of bandwidth (in bit/s) while, at the same time, minimize the total number of active wavelengths (i.e., minimize the power consumption at the OLT, minimizing the number of line cards used). Moreover, in our scenario the ONUs connected to the ring network provide traffic belonging to different service (e.g., fixed traffic, 3G, LTE, ...) with different requirements in terms of bandwidth. Since different types of traffic are managed through different network protocols and different network nodes, we add a constraint to our ILP that groups the transmissions of the ONUs providing the same type of traffic over the same wavelength (or the same set of wavelengths). We argue and we aim at demonstrating that physically separating the different types of traffic over different wavelengths provides a benefit in terms cost of traffic management (e.g., processing delay in the nodes). With this constraint, the outcome of the ILP is a set of Virtual-PONs (VPONs). VPONs are PONs where the ONUs belonging to a particular VPON are not necessarily connected to the same passive remote node. The ONUs belonging to a VPON can be connected to different branches of a PON or LR-PON (e.g., connected to different passive remote nodes of a ring optical access network), and their transmissions are physically separated from the transmissions of the other ONUs by using a dedicated wavelength (or set of wavelengths). The allocation of the transmissions of a VPON is managed by a single OLT. For this study, an ILP has been developed and solved with CPLEX. The results of this research will be of impact for the future generations of high-speed broadband and mobile network infrastructure. An FMC access network will lead to significant network cost and energy-consumption reductions which will be key to address the profound transformations needed to face data traffic explosion in the medium to long term.