Time is of the essence. We say that someone is reliable if he is
punctual. We say that an enterprise is reliable if, in the face of a
problem, they try again. On the other hand, most commercial contracts
contain penalty clauses to deal with ;SPM_quot;late delivery;SPM_quot;.
Thus we can see that timeliness and reliability are inherently bound
together. This chapter is about the mechanisms that an Open
Distributed System must provide to ensure some required
level of reliability and timeliness. In practice this means providing
fault and performance transparency for each user or application.
This is in contrast to the timeliness required by multimedia
communication.
Firstly, we overview some of the standard reliability techniques
appropriate to a distributed system. Then we look at the special
problems of shared resources in such a system. The we look at how to
provide higher reliability using transparent replicated objects.