Browse Prior Art Database

Transparent continuation of controller applications in redundant networks

IP.com Disclosure Number: IPCOM000010844D
Original Publication Date: 2003-Jan-27
Included in the Prior Art Database: 2003-Jan-27
Document File: 4 page(s) / 13K

Publishing Venue

IBM

Abstract

Applications running in a network of embedded controllers require a high degree of transparency with respect to the actual network topology. When failures occur in the network domain, for instance, a network interface adapter fails, the software associations with this interface inside the TCP/IP stack are invalid. Applications will experience a disruption of service in this case. We present a method of transparently continuing these applications in the presence of such network failures if the physical network provides redundant interfaces at each controller station. The "swap link" method rmaps the internal resources to another interface and transparently routes the network packets over the redundant network.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 34% of the total text.

Page 1 of 4

Transparent continuation of controller applications in redundant networks

Problem statement

    In the world of TCP/IP applications use the familiar BSD socket API that implicitly defines the notion of "endpoints" to describe the network resources used by communicating application at a time. These endpoints are five-tuples consisting of two IP addresses, two port numbers and a protocol specification. If network failures occur while these applications are executing, their endpoints will be destroyed. Applications required to close and reset the communication at both ends.

Principles of the swap link solution

    Therefore it is a highly desirable goal to transparently continue communicating applications in the presence of link failures. This goal can only be achieved if the endpoints that are maintained on behalf of the applications inside the IP stack can be reused or never experience any disturbances.

    The idea of a transparent continuation is pursued on the lowest level of networking protocols that operate "behind the curtain of addressing". If the addresses that are in use by an end-to-end communication can be reused after a recovery action then a maximum degree of transparency can be achieved. The endpoints used by an application can be preserved.

    The proposed solution will be called the "swap link" solution. The basic idea is derived from a network structure that connects all network interfaces of controllers to the same shared network media in the sense that any station can receive all Ethernet frames from all other stations. In this network one can re-assign network addresses from one interface to another one without disabling the use of these addresses by other stations.

Network structure and configuration

    The basic idea is extended to a network structure that is composed of two physically distinguished networks implemented as Ethernet hubs or switches. Controllers featuring at least two physical network interfaces are connected to these hubs in a symmetric fashion. This structure avoids any single points of failures (SPOF) by decoupling the hubs and connecting all controllers to each of them. Figure 1 shows the basic structure for a set of controllers named A,B,C and R.

    The configuration assumes that network ZERO is operated as the master or main functional network, and that network ONE is operated as a stand-by network (but the method can be extended to configuring network ONE also). In case of link failures occurring in network ZERO, a link from network ONE can be attached to the main network.

    1. The two networks will be used and configured in an asymmetric fashion. Network ZERO (0) shall be named the "functional" or "master" network. It will be used by communicating applications running distributed among the controllers. Network ZERO shall be configured as a standard IP network using a defined, probably reserved set of IP addresses such that addressability between controllers in any-to-any mode is given. The sample network in F...