Abstract
We describe an approach to construct a middleware subsystem with varying levels of security and reliability, from tolerance to simple transient faults at one end to tolerating malicious attacks and intrusions at the other end. Our approach involves two steps. First, process duplication and mutual-checking are used to construct failsilent processes which are guaranteed to fail only in a crash manner. Secondly, middleware processes (to be duplicated) are programmed to implement crashtolerant protocols. The proposed, two-tier approach is effective if: (i) the chosen crash-tolerant protocols are deterministic, (ii) there is one replica within a given replica pair which functions correctly until it observes the failure of the other replica. We also estimate the redundancy required for constructing a fault-tolerant replicated service over an asynchronous network (e.g., the Internet) using our approach to building middleware.
A Two Tier Approach To Building Dependable Middleware Services
In Workshop on Dependable Middleware-Based Systems; part of the IEEE International Conference on Dependable Systems and Networks (DSN 2002), Washington DC, USA, 23-26 June 2002
pp G49-G54
IEEE Computer Society Press, 2002
[Abstract]