Title: The Reliable Hybrid Pattern - A Generalized Fault Tolerant Software
       Design Pattern

Authors: Fonda Daniels (Dept. of Electrical & Computer Engineering, Box 7911)
         Kalhee Kim and Mladen A. Vouk (Dept. of Computer Science, Box 8206)
         North Carolina State University, Raleigh, NC 27695

Contact author: Fonda Daniels
                 Dept. of Electrical & Computer Engineering, Box 7911
                 Raleigh NC, 27695-7911
                 919-515-7479
                 email: fjdaniel@eos.ncsu.edu

Key Words: Reliable hybrid pattern, fault tolerant software, N-Version
Programming (NVP), Recovery Block (RB), Consensus Recovery Block (CRB), Maximum
Likelihood Voting (MLV), Consensus Voting (CV), Acceptance Voting (AV), N-Self
Checking Programming (NSCP), Expedient Voting (EV)

Abstract

In order to make object-oriented software applications more reliable, it may be
necessary to incorporate into them a strategy for tolerating software faults.
In the case of critical applications, this is essential. In this paper we
describe a general pattern, which we call the Hybrid pattern, that can be used
to design fault tolerant software applications.  The pattern supports
development of applications based on classical fault tolerant strategies such
as N-Version Programming (NVP) and Recovery Block (RB), as well as those
based on advanced hybrid techniques such as Consensus Recovery Block (CRB),
Acceptance Voting (AV), N-Self Checking Programming (NSCP) and Expedient Voting
(EV).  The Hybrid pattern reflects that these advanced strategies can be
constructed through recursive combination of NVP, RB, and hybrid components.
Our pattern also explicitly indicates how different advanced software voting
techniques fit into the design structure.  We use CRB to illustrate the
behavior of the designs that can be constructed using the Hybrid pattern.