Reliability of networked applications

The aim of this article is to present the dependability of networked applications from a practical angle. We begin by briefly describing the concepts of dependability in computer systems, then focus on one of the means of achieving it - fault tolerance - which is essential if the system is to continue to function despite the presence of faults that cannot be eliminated or prevented. We then give an overview of the techniques that can be employed, focusing on the particularities of distributed systems, demonstrating the complexity of developing fault-tolerant network applications.

The last two parts of this article illustrate the practical application of the techniques presented above. First, we mention clustered architectures, for which computer manufacturers provide numerous turnkey solutions. We then describe the recent FT CORBA (Faut-Tolerant Common Object Request Broker Architecture) standard from the OMG (Object Management Group), recommended to support software publishers for distributed object-based applications, and discuss recent work by the IETF (Internet Engineering Task Force) on Internet applications. We conclude with a case study on the reliability of an Internet router.