Ion Popescu

On the optimal design of a spectrum-switched optical network with multiple modulation formats and rates

Spectrum-switched optical networks (SSONs) are an attractive solution for core networks. With respect to wavelength-switched optical networks (WSONs), they achieve a higher spectrum efficiency thanks to the use of a flexible grid, instead of a fixed one. In SSONs, the well-known wavelength assignment (WA) problem of WSONs becomes a spectrum assignment (SA) problem. The SA problem aims at assigning a portion of the spectrum (called a frequency slot) to lightpaths. The width of the frequency slot depends on the requested bit rate and the modulation format adopted for the lightpath, and it is a multiple of the minimum bandwidth granularity, referred to as frequency slice. Thus, differently from WA, which assigns a single wavelength (or color) to each lightpath, SA assigns a set of colors (i.e., a set of frequency slices) to the lightpath. The shift from WA to SA introduces an additional constraint, which is related to the spectral adjacency of such frequency slices. This paper proves that the adjacency constraint in SA is not required and that by solving the WA problem (or the coloring problem) it is possible to derive a solution with spectrally adjacent slices in polynomial time. Based on such results, an integer linear programming formulation (ILP) for the optimal SA in a SSON with multi modulation formats and multi line rates (MMF/MLR) is presented for minimizing the network cost. The total cost of the network comprises the spectrum cost and the transponder card cost. Optimal results are presented for a MMF/MLR-SSON ring and show the amount of total occupied bandwidth and the network costs for different loads and ring lengths. Optimal selection of the modulation format and line rate is driven by the slice cost and the transponder card cost. Result comparison indicates that support of MMF and MLR is especially effective for improving spectrum utilization and allows spectrum saving up to 20% with respect to an SSON ring with single modulation format and line rate.

A cost comparison of survivable subwavelength switching optical metro networks

In this paper, we compare the cost of the survivable optical transport solutions with subwavelength switching granularity. The focus is on Time-Domain Wavelength Interleaved Network (TWIN) and Packet Optical Add/Drop Multiplexer (POADM) network. Both technologies exploit optical slot switching: TWIN on a physical mesh topology, and POADM on a physical ring. The cost model accounts for the node components, the transponder cost and the cost of wavelength use per km of fiber. For the first time, a single network dimensioning model supporting protection is proposed for these metro technologies. The network planning solution is based on linear programming, and addresses the routing and wavelength assignment problem (RWA) and the optimal scheduler calculation. We consider dedicated protection, which consist in doubling the capacity of each operational connection network and shared protection, in which the protection capacity is potentially shared between multiple backup flows. By running the linear programming optimization we study the impact of different protection models on the cost of networks in different configurations and identify the scenarios for which each of the technologies is best suited.

Software-defined optical networks and network abstraction with functional service design [Invited]

The combination of software-defined networking (SDN) and optics, i.e., a software-defined optical network (SDON), offers combined benefits that include flexible SDN control and large optical transport capacity. In this paper, we introduce the unified SDN as a common control solution for end-to-end OCS/OPS networks, with the introduction of an additional network element, namely the OpenFlow agent. The SDN control plane is implemented in the OPS network for simplified data plane design, optimal network organization, and contention alleviation, as well as improved control flexibility and advanced handling of network dynamics. The SDON demonstration is conducted in the nationwide transport networks constructed with different technologies via various internetworking testbeds. Based on the SDON platform, we further introduce the service design concept for advanced network abstraction. The general network architecture based on a service design concept offering flexible, reconfigurable, and economical end-to-end (FREE) service features is described. This service design aims to provide customized service and maximum network freedom with virtualized and flexibly selected functionalities, while maintaining a moderate network configuration cost and promoting network innovation at the same time. Moreover, the service design concept could easily be extended to any layers and any network domains, thus providing the unified network abstraction ability and the end-to-end control solution. Based on the service design, control, coordinated, and orchestrated optical transmission is demonstrated by reconfiguring the virtualized optical functions. This type of approach to network abstraction is further exploited in the heterogeneous multidomain networks, where differentiated and featured service provisioning is demonstrated.

Synchronization of the Time-domain Wavelength Interleaved Networks

We have addressed synchronisation of optical packet switched network on a mesh topology. The focus here is the synchronisation architecture for the optical packets in TWIN based network. The optical packets from different sources are passively groomed on one wavelength with a central control/management plane. In order to provide the necessary precise synchronisation system between the central control entity and all other nodes the synchronisation information time-stamp is carried on in band dedicated control channel throughout the network. In addition, the synchronization mechanism enables the network to operate in presence of realistic temperature variations of the fibre and include a correction of the effect of chromatics dispersion on the propagation delay calculation.

Optimal and dynamic virtual datacenter provisioning over metro-embedded datacenters with holistic SDN orchestration

Abstract Conventional mega datacenter (DC) lacks flexibility and has poor service connectivity for users faraway. Therefore, the metro-embeded DC architecture has been introduced to deliver more reliable and flexible services with less access latency. For fine-grained all-optical switching ability and high bandwidth connection, optical burst switching (OBS) ring is deployed inside the micro-DCs (mDC) and OBS over wavelength switched optical network paradigm is used for inter-mDC communication. Meanwhile, hierarchical SDN control is employed to virtualize and holistically coordinate the distributed mDCs and the metro network slices into reconfigurable virtual DC (VDC). Optimal VDC provisioning based on an integer linear programming (ILP) formulation is also developed in the SDN control plane to provide services with minimum latency and traffic load. A dynamic VDC provisioning algorithm based on partial ILP is further proposed to reconfigure the VDC with reduced complexity considering the user and resource dynamics. Experimental demonstrations conducted on the implemented metro-embedded DC prototype testbed show that optimal and fast VDC provisioning is realized based on the proposed ILP, while dynamic VDC reconfiguration makes sure that VDC is adjusted accordingly with the change of user location and resource requirement.

Scalable routing, scheduling and virtualization for TWIN optical burst switching networks

We propose a comprehensive scalable algorithm simultaneously assessing the routing, scheduling and virtualization in Time-Domain Wavelength Interleaved Network (TWIN). TWIN is an energy-efficient optical burst switching technology for metropolitan area and data center intra- and interconnections, with destinations using separate allocated sets of wavelengths for the reception. Given the costs of the optical transponders and the wavelength use per km of fiber length, the proposed algorithm solves the routing and wavelength assignment problem, and in the same time allocates the time slots to traffic flows (performs the slot scheduling), so that the total network cost is minimized. The algorithm also enables the construction of overlayed virtual networking domains at minimum cost, which share the transponders and network links, and can employ different scheduling policies. The performance of the algorithm is compared with the pre-existing optimal dimensioning solution for single virtualization domain based on Integer Linear Programing, for different scenarios. The obtained results show that the network cost is within 27% of the optimal.

Cost of protection in Time-Domain Wavelength Interleaved Networks

In this paper, we study the cost of protection in Time-Domain Wavelength Interleaved Network (TWIN), an optical burst switching technology for transport networks with active edge nodes, passive core nodes, and time slotted operation. For the first time, an optimal network dimensioning with protection is proposed for TWIN, allowing to allocate the needed resources for dedicated and shared path protection. The dimensioning model is in the form of a 0-1 Integer Linear Program (0-1 ILP), and addresses both the routing and wavelength assignment problem, and the scheduler calculation. The cost of protection is studied for centralized and distributed traffic models, and it is shown that the shared path protection in TWIN is the most cost efficient.

Virtual ring based protection for Time-Domain Wavelength Interleaved Network

We propose a protection scheme for Time-Domain Wavelength Interleaved Network (TWIN), an optical burst switching technology for transport networks, where each destination node uses a dedicated wavelength, while sources are equipped with fast tunable lasers. TWIN is composed from several multipoint-to-point trees built on top of a mesh network linking sources and destinations. Each multipoint-to-point tree in TWIN has its root in a destination node, and is thus used for the reception at a single, unique wavelength. In the proposed method resource allocation is centrally performed in order to avoid contention on the multipoint-to-point trees while optimizing source activity. We propose using a virtual ring joining all sources and destinations in order to support both an out-of-band control channel and all protection paths. The control channel may be used in particular to carry messages related to network discovery, synchronization and protection issues. Synchronization and failure notifications (generated by standard optical layer monitoring procedures) are processed by the central element in charge of resource allocation. The proposed TWIN layer protection approach avoids using lower layer protection techniques, which may strongly impact the fiber length distribution, and thus the associated scheduling mechanism, inside the different multipoint-to-point trees. The use of a single virtual control ring is well suited to networks with a limited extension such as metro networks. Our approach can however be extended to larger scale networks by using several control rings. We describe an algorithm to cover a large network operated by TWIN with minimum number of interconnected rings, taking into account the maximum transparent reach for protection paths.

Carrier-grade performance evaluation in reliable metro networks based on optical packet switching

In this paper we assess the performance delivered in a metro network by two optical packet switching architectures, enabling sub-wavelength switching granularity. We compare POADM (Packet Optical Add/Drop Multiplexer) with TWIN (Time-domain Wavelength Interleaved Network). These technologies are envisaged to be deployed in the metropolitan area in order to improve bandwidth utilization and minimize energy consumption, thanks to their excellent switching granularity. In order to perform a realistic performance assessment, the study was carried out over a large set of traffic demands, for which the candidate architectures were first dimensioned taking into account network reliability. Then, the delivered data plane performance was assessed by simulation in terms of electronic packet loss, jitter and insertion delay. Two routing scenarios (any-to-any and hub-and-spoke) are mapped on a physical topology inspired from a European operator network. Simulation results show that both architectures easily achieve performance targets set by the Metro Ethernet Forum, as long as the network is properly dimensioned.

Adaptive DC interconnection provisioning in distributed all-optical micro-datacenters using holistically SDN orchestration for dynamic access

The metro-embedded datacenter (ME-DC) architecture which is based on micro-DCs (mDC) is introduced and implemented to deliver more flexible services with less access latency. Hierarchical SDN control is employed to adaptively interconnect the distributed mDCs and the metro network slices into re-configurable virtual DC (VDC). Such an approach improves the overall resource mobility, especially for traveling users that need to dynamically access the service. In this paper, the VDC re-configuration and pre-configuration mechanisms for dynamic user access are proposed and demonstrated for the first time. In specific, by re-configuring the VDC with the most accessible resources with user location tracking, fast and flexible service is provided with consistent service access. Furthermore, by pre-configuring the VDC with user prediction, mDC and metro network resources are reserved in advance for users in order to provide seamless service. Demonstration based on the implemented ME-DC prototype shows that VDC is adaptively constructed for dynamic user access with fast and flexible service provisioning.