create new tag
, view all tags

Experimental Network Infrastructure

Optically Switched Core

As part of the DRAGON project, we are building a optically switched network in the Washington D.C. metropolitan area. The purpose of this experimental network is to develop and demonstrate the capabilities of a "hybrid" packet and circuit switched network infrastructure.

DRAGON Optical Core as of 2005-12-19


[ jpg ]
DRAGON Architecture


The DRAGON Architecture is similar to contemporary R&E networks in the context of topology and administrative organization. It consists of a number of autonomous network domains, each able to unilaterally define internal traffic management policy, and enter into bi-lateral peering arrangements with other external domains where desirable and mutually agreed. The fundamental difference is that the DRAGON Architecture is based on switching and forwarding nodes that support the GMPLS label hierarchy

The GMPLS label hierarchy comprises packet labels (e.g. MPLS), TDM labels (e.g. SONET), wavelength labels, and fiber labels. Switching nodes that operate on traffic according to GMPLS labels are referred to as Label Switch Routers (LSR). An LSR is not required to support all GMPLS label types; instead it will support a subset based on equipment and network capabilities.

The DRAGON Architecture further assumes that each LSR runs an intra-domain GMPLS routing protocol (e.g., GMPLS-OSPF) and a GMPLS path signaling protocol (e.g., GMPLS-RSVP). Alternate routing and signaling protocols may also be used. The combination of label switched routers and associated signaling allows for the dynamic control and instantiation of Label Switched Paths (LSPs) which provide end-to-end connections for multi-service traffic at the transport level consistent with the specified label type.

Intra-Domain establishment of lambda based LSPs is reasonably mature due to standardized extensions to the Interior Gateway Protocols (IGP) and signaling protocols. However, inter-domain instantiation of end-to-end LSPs which can be dynamically provisioned across multiple administrative domains and heterogeneous network technologies is still problematic. As depicted in Figure 1, there are several capabilities and technologies missing in order for this capability to be readily available to end systems; i) no standardized inter-domain routing architecture for LSPs, ii) no simple application interface, iii) no end to end instantiation (with proper authentication, authorization, and accounting), and iv) no ability to signal thru non-GMPLS enable network segments.

We propose to develop an architecture for inter-domain end-to-end GMPLS transport. This DRAGON Architecture will provide the missing components in the form of Network Aware Resource Broker (NARB), Virtual LSR (VLSR), and an Application Specific Topology Definition Language (ASTDL):

  • In order to establish end-to-end inter-domain LSPs, interconnected networks must exchange information corresponding to the label types each offers, which destinations can be reached via which label types, and which users are allowed to use which label types. We propose the concept of a Network Aware Resource Broker (NARB) that represents an autonomous domain and exchanges this information with NARB instantiations representing other domains. It is this inter-domain exchange that enables end-to-end LSP routing. An integral part of this exchange will be authentication, authorization and accounting information.

  • We propose the concept of a Virtual Label Switch Router (VLSR) to translate standard GMPLS protocols into device specific protocols, to allow dynamic reconfiguration of non-MPLS aware devices. This is a pragmatic issue in that it allows non-GMPLS switching devices to be managed via SNMP or an other proprietary protocol and still be an integral part of the end-to-end GMPLS environment.

  • We propose to define an Application Specific Topology Definition Language (ASTDL) to formalize network service definitions and to simplify description of complex network topologies. We also propose to develop tools and middleware libraries to interface user applications with the routing, signaling, and security framework of the proposed DRAGON Architecture.

This architecture will be instantiated in the Washington DC metropolitan area, building on our previous experience with operation of the MAX GigaPOP and other regional testbed networks.

Additional information can be found in Policy-Based Resource Management and Service Provisioning in GMPLS Networks.

Dark Fiber

Fiber Distances Between DRAGON Sites

Full text: PDF

Fiber Distances Between DRAGON Sites

Color Type Provider
Purple Allwave Qwest
Green SMF-28 FiberGate
Aqua SMF-28 Level3


Real-time Performance Monitoring

Cricket Performance Monitoring

Nagios Network Health Monitoring

Mailing Lists

Topic revision: r11 - 2014-04-07 - XiYang
This site is powered by the TWiki collaboration platform Powered by Perl Copyright 1999-2021.
The information contained in these pages is the property of the Mid-Atlantic Crossroads (MAX).
If you have questions or comments, please contact WebBottomBar">MAX Administration