GMPLS Control Plane Emulator with PCE
| |
| ADRENALINE Test-bedĀ®: GMPLS-based Control Plane Emulator with PCE |
Contents
|
GMPLS-enabled controllers
From a research point of view, one of the focus and goals of the ADRENALINE testbed is the performance evaluation of GMPLS-based traffic engineering algorithms and schemes. For this purpose, a new set of 74 GMPLS-enabled controllers without associated optical hardware (i.e., the optical hardware is emulated) has been added. This set of GMPLS controllers introduces a new degree of flexibility in topology configuration, without restrictions regarding the targeted optical network topology or regarding the resources per link (e.g., number of available wavelengths, fibers, etc). Thus, the GMPLS controllers can be inter-connected following any devised topology, by means of Ethernet point-to-point channels carried over emulated optical links. The proposed solution allows the specification of control link parameters for realistic QoS constraints (fixed and variable packet delays, packet losses, bandwidth limitations, etc.). In particular, and in order to provide a flexible framework for topology reconfiguration, the IP Control Channels (IPCC) in the DCN are point-to-point IP interfaces over Ethernet interfaces. To do this, it uses virtual local area networks (IEEE 802.1q VLANs), configured both in the layer 2 Ethernet switches and in the GMPLS-enabled controllers within the testbed, with optional GRE or IPIP tunnelling. This approach enables the deployment of arbitrary layer 2 interconnections between network nodes absolutely decoupled of the physical infrastructure. Finally, let us note that the set of 74 GMPLS-enabled controllers are implemented in Linux-based routers with Intel Xeon 3.0GHz, Core 2 Duo E6550 2.33GHz and Intel Core 2 Duo E8200 2.66GHz processors.
Virtualized GMPLS-enabled controllers
Both the KVM (Linux Kernel based Virtual Machine) and the Xen virtualization software techniques have been introduced in the testbed, allowing to build virtualized GMPLS-enabled controllers and emulated GMPLS networks in just a single Linux-based physical host. Both techniques are currently being used, considering different factors such as the presence of hardware virtualization technology on the host CPUs and the virtualized I/O. These virtualized GMPLS-enabled controllers run the same set of software as the rest of GMPLS-enabled controllers, also with emulated optical hardware. The virtualized network provides a cost-efficient solution, leveraging the benefits of server consolidation. Such virtualized networks are normally used for prototyping and extension development. The virtualized GMPLS-enabled controllers are configured in the same way as the actual GMPLS-enabled controllers, and other than the performance bottlenecks with a high number of guest systems, they are functionally equivalent to the emulated ones. Currently, we have deployed 2 virtualization servers with Intel Core 2 Duo E8200 2.66GHz processors and 4GB of RAM.
ADRENALINE Network Configurator (ADNETCONF)
|
The flexibility of ADRENALINE control plane emulator allows the deployment of as many different network topologies and configurations as desired. However, this flexibility usually comes at the expense of high management costs when passing from one scenario to another one, as well as extreme time-consuming, and error-prone manual procedures. These problems can be overcome through implementing automatic reconfiguration procedures such as the use of ADNETCONF (ADrenaline NETwork CONFigurator), a software application in charge of scenario model management in ADRENALINE testbed. It has been developed with two main objectives in mind. First, to provide an easy, quick and intuitive way of designing scenario models and secondly, to implement automatic deploy, undeploy and monitor operations of such models. The application is composed of two main modules, a graphic user interface (providing the model editor, in addition to the interface to launch operations, Fig. 1) and a processing engine (implementing the actual operations). |
ADRENALINE Network Generator (ADNETGEN)
|
Additionally, the testbed also includes a tool for dynamic statistical generation of lightpath connection requests in the optical network, named ADNETGEN (ADrenaline NETwork GENerator). This GUI-based tool (Fig. 2) allows researchers and network operators to define the traffic patterns: statistical distribution (e.g. uniform), total number of connections and its inter-arrival and duration (i.e. holding) times. ADNETGEN is also in charge of logging the outcome of each connection (i.e. successful or failed establishment, average setup delay, etc.) and providing final results once the experiment is finished. |
Control Plane Interconnection
Scope and Motivation
- Research on Multi-Domain network architectures
- Research on Multi-Region Networks (MRN) and Multi-Layer Networks (MLN)
- Collaborative / Integrated research involving several test-beds
- Proof-of-concept scenarios, architecture validation
- Interconnection
Common Use Cases for ADRENALINE testbed interconnection
Multi-AS WSON
Multi-Area WSON
Multi-Layer scenario : connection Oriented Ethernet over WSON
Technical Details
Success cases
In the scope of research projects, we have successfully interconnected the GMPLS Control Plane Emulator of ADRENALINE testbed with the third party testbeds
|
|
CELTIC TIGER2 project
CELTIC 100GET project
RAION2 project
FP7 ICT NoE BONE project
|
FP6 IST IP NOBEL2 project
|
EUREKA ITEA TBONES project
|