Fig 2. Logical architecture of ADRENALINE all-optical nodes: a) Developed Optical Cross Connect (OXC). b) Developed Reconfigurable Optical Add Drop Multiplexer (ROADM).

The ADRENALINE testbed is composed of an all-optical Dense Wavelength Division Multiplexing (DWDM) mesh network (Fig. 1) with two colour-less Reconfigurable Optical Add Drop Multiplexer (ROADM) nodes (Fig. 2.b) and two Optical Cross Connect (OXC) nodes (Fig. 2.a), providing reconfigurable (in space and in frequency) end-to-end lightpaths, transparent to the format and payload of client signals (e.g., SONET/SDH, Gigabit Ethernet). Each optical node has two DWDM transceivers up to 2.5 Gb/s and one at 12.5 Gb/s with fully tuneable laser sources. The optical node architecture (Fig. 2) is based on using Array Waveguide Grating (AWG) as DWDM (de-) multiplexers (8 and 16 wavelengths with 50 and 100GHz channel spacing, respectively), and Micro Electro Mechanical Systems (MEMS) as the switching technology. Arrays of power meters and Variable Optical Attenuators (VOAs) are used for optical power equalization at output fibers. ADRENALINE deploys a total of 610 km of G.652 and G.655 optical fiber divided in 5 bidirectional links, in which optical amplifiers (Erbium-Doped Fiber Amplifiers or EDFAs) are allocated to compensate power losses during optical transmission and switching at C-band. ADRENALINE transport plane also includes non-intrusive Optical Performance Monitors (OPM) to obtain spectral information tapping a 5% of all the input and output fibers, namely, channel and in-band Optical Signal to Noise Ratio (OSNR), channel and aggregate optical power, and wavelength drift.

Fig. 3. Logical architecture of ADRENALINE’s GMPLS-enabled Connection Controller

Each optical node is equipped with an Optical Connection Controller (OCC) for implementing a distributed Generalized Multiprotocol Label Switching (GMPLS)-based distributed control plane [IETF RFC3945]. Such a control plane is responsible for handling, dynamically and in real-time, the resources of the optical node in order to manage the automatic provisioning and survivability of lightpaths (using the Resource Reservation Protocol – Traffic Engineering (RSVP-TE) signaling protocol for wavelength reservation, and the Open Shortest Path First – Traffic Engineering (OSPF-TE) routing protocol for topology and optical resource dissemination), allowing traffic engineering algorithms with Quality of Service (QoS). The GMPLS-enabled Controller executes several collaborative processes that are in charge of tasks such as the management of optical resources (Link Resource Manager or LRM); hardware abstraction layers and drivers (Hardware Controller); protocol entities such as the routing controller with OSPF-TE, the connection controller with RSVP-TE or the Link Management Protocol or LMP process for control channel connectivity/correlation and fault localization and, finally, SNMP agents for SNMP based network management and dedicated Path Computation Elements (Fig. 3), the system running in a Linux-based router with an Intel Core 2 Duo E6550 2,33 GHz processor. The Data Communication Network (DCN) is based on IP control channels (IPCC) carried at 1310 nm and C-band at the optical fiber with a line rate of 100 Mb/s using point-to-point links.

Fig 4. Functional architecture of the Path Computation Element (PCE)

The IETF PCE working group has defined a Path Computation Element as an entity (component, application or network node) that is capable of computing a network path or route based on a network graph and applying computational constraints. It involves a (standard) functional “formalization” of a network architecture and a communications protocol interface (PCEP). Additionally, new collaborative approaches and hierarchical models are being proposed for advanced (multi-domain and multi-layer) path computation solutions. It is worth noting that the use of a PCE does not imply a centralized path computation (several PCEs could be deployed in a signle domain, including one or more PCEs per network node) and does not preclude hybrid-mixed deployments. It does not replace a Network Management System, but may be seen as a functional part of it.

The ADRENALINE network includes a Path Computation Element (PCE), which is a dedicated network entity responsible for doing advanced path computations. The PCE serves requests from Path Computation Clients (PCCs), and computes constrained explicit routes (EROs) over the topology that constitutes the optical transport layer. The selected PCE deployment model is based on deploying a single PCE per OSPF-TE area, co-located in a GMPLS-enabled Controller node and coupled to a Routing Controller. The preferred synchronization mechanism, by which the PCE constructs a local copy of the Traffic Engineering database (TEDB) is non-intrusive: by sniffing OSPF-TE traffic, it constructs a dedicated (i.e. not shared) database using stateful inspection of the TE Link State Advertisements (LSAs) contained within the OSPF-TE Link State (LS) update messages, thus passively reusing the OSPF-TE dissemination mechanism, and not requiring the creation of an additional listener adjacency.

The functional architecture of the PCE is shown in Fig. 4. The PCE is a multi-threaded, asynchronous process, serving requests in a client/server approach. Upon acceptance of a client connection, the Finite State Machine drives the PCE Communication Protocol (PCEP) protocol and, after the initial handshake, PCCs may send Path Computation Requests. A dedicated thread is responsible for updating the Traffic Engineering database, while a thread pool is used for the actual path computation, using a writer/readers lock. Algorithms are implemented in shared libraries, using an abstracted algorithm Application Programming Interface (API). This allows algorithm implementations to consider the network topology and the Traffic Engineering database as a directed graph, and to request path computation to other PCE peers for distributed or collaborative Path Computation.

The ADRENALINE PCE Server is a network appliance that implements a Path Computation Element, detailed in Path_Computation_Element_(PCE).