Special Purpose IP Routing, 10-R8243
Inclusive Dates: 07/01/11 – 12/31/12
Background — Network solutions permeate all domains, from our homes and cars, to industry and space. The success of networking stems from the growth and standardization of the Internet Protocol (IP), which provides a universal open transport. The openness of this protocol makes it such that multiple vendors have fielded interoperable devices. Almost every company has network closets filled with blue boxes (Cisco® routers) and every home has a network connection that includes a router and wireless access point. This openness in standardization comes with a cost. Special purpose and flexibility beyond the envelope of support is very hard to reach. For example, simply adjusting a home router to better handle overload in the Voice over IP scenario is a serious challenge. Consequently, most networks are only used in their default settings with performance accepted and the functionality left as a mystery. The ubiquitous nature of networks has led to a desire to leverage networking technologies in special communications arenas. Unfortunately, not all of these scenarios are well served by “default-setting” networking. At times, for the small volume specialized scenarios, the existing vendors are unwilling or unable to adapt or adjust their products. For example, Cisco has no business interest in providing specialized product markets smaller than 100s of millions of dollars.
Approach — A framework was developed for rapid development and deployment of special purpose IP routers. This was achieved by adjusting settings on standard routers (rather than using the typical “default settings”) and combining with embedded computer nodes. The embedded computer nodes can leverage knowledge about the inner workings of the commodity devices and manipulate data flows such that the overall specialized scenario needs are achieved. The approach to this project was to research, define and characterize the virtual router concept and to evaluate its suitability to solving the types of problems similar to flight test telemetry networks. This required evaluating open-source router source code to determine portions that can be leveraged for creating the router virtualization and adding the distributed concepts necessary to implement a virtual router. A series of challenge problems was developed to evaluate the virtual router concept. Baseline performance was measured using standard routers with both default and optimized configurations. The virtual router was evaluated using the same challenge problems and compared to the baseline performance. The virtual router implementation was successively revised and reevaluated to determine performance gain over baseline.
Accomplishments — A number of open-source routing packages were evaluated to determine which looked most promising for a starting point in developing the virtual router concept. Standard and modified router implementations were built based on several of these packages. The virtual router architecture was augmented using distributed physical network interfaces under the control of a central control node. A combination of open source software capabilities not originally designed for routing applications was leveraged to enhance the scalability and flexibility of the virtual router approach. A laboratory environment was established by implementing multiple virtual router nodes using a combination of SwRI network lab resources. The main test bed setup consisted of nodes distributed within an SwRI building, inside a residence in San Antonio, and in a hotel room in Pennsylvania. Network performance was measured across these three networks under varying network load conditions both with and without the virtual router. The project demonstrated that with a few modifications in limited locations to existing networks better control of network data can be maintained through the existing network infrastructure, allowing for improvements in efficiency and performance between congested networks.