SAN FRANCISCO A software resource layer and tool set introduced by Eternal Systems Inc. at this week's Embedded Systems Conference here is designed to help build a fault-tolerant embedded infrastructure.
Eternal Systems (Santa Barbara, Calif.) said the system automates the common practice of system state check-pointing used to replicate application code and restart it with a minimum of down time after a system failure.
Robert Monkman, director of marketing, said the new software infrastructure layer, called “Duration,” is aimed at applications that require the benchmark “five-nines” (99.999 percent) reliability or greater levels of service availability. While optimized in this initial offering for use with Linux, the software is platform, OS and Web services framework agnostic.
The technology's introduction comes as a growing portion of embedded developers are focused on mission-critical market segments such as telecommunications, defense, transportation and industrial control.
Most commercial proprietary real-time OSes have been enhanced to improve their ability to support fault-tolerant operation in high availability networking applications. Linux vendors such as MontaVista and Red Hat have done likewise.
A variety of manual check-pointing tools from companies such as Wind River Systems and Go-Ahead provide increased reliability to the applications that run on and between the networked devices.
Three unique features of Duration, Monkman said, are automation of the checkpoint process, the smaller amount of additional code and programming effort and the underlying hardiness of the software replication process.
“The automated check-pointing engine we have built as a part of our software suite essentially links in underneath the application, layered between the application and the underlying hardware and OS,” said Monkman.
At the other end of the spectrum is active replication, most often used in large enterprise and mission-critical applications where processing, memory and resources are available to allow each system state to be updated every time there is a change.
Eternal said it uses a semi-passive replication and check-pointing approach. System messages are queued up and saved on a continuing basis. This information is then combined with the previously saved replica and state information to resynchronize the state of the system applications.