Automated and optimised heartbeat function for IP connections

Automated and optimised heartbeat function for IP connections

This article explains the background to one of the problems of monitoring and maintaining connections over an IP network. It indicates the components that software systems use to access such a connection. It outlines how such components can be optimised so that they impose a minimal impact on network traffic by the use of a simple mechanism to achieve their aims.

    A connection is a path through a network between a pair of nodes. Computer systems that communicate with each other over a network act as pairs of network nodes. From the point of view of each of these nodes there is a logical connection between them allowing them to send and receive messages with each other. The path spanned by a single connection may pass through other network control devices, and in the case of an IP network these may be routing components called firewalls. The firewall can apply network management controls to the connections that they have access to. Typically the set of controls that a firewall enforces include the ability to block invalid connection requests or to terminate existing connections that have remained unused for specific periods.

    The computer systems that access these networks generally contain some network management component that is used to establish the connection between the nodes, to handle individual messages that are sent and received over the connection, to release the connection when it is no longer needed, and to process a range of error conditions that may be encountered during these operations. Additional components in each computer system manage a set of separate programs that issue messages to programs running in a partner system, or which are attached in response to the arrival of messages from a partner, and together these programs constitute or form part of the workload that each computer system is responsible for.

    The frequency of messages being transmitted between pairs of nodes can vary considerably and there may be prolonged periods where little or no network traffic takes place. It can be difficult for the network controller components to be made aware of problems that occur during periods of network inactivity, as errors are unlikely to be detected until further messages are transmitted between the network nodes and subsequently fail. Such network failures can interrupt the successful operation of a workload and may then require manual intervention to rectify the situation. In addition the complexity of networks and the lack of controls within firewalls often make it difficult for customers to effectively manage wide are network connections across their enterprises.

    The network management component of a computer system can maintain a record the successful receipt of messages passing over each of the connections that it is responsible for, and uses this information as an indication that a particular connection is still functioning correctly. The network...