Asynchronous Recoverable OLTP with Graceful Degradation

I am looking for references, papers, etc. to the theoretical, practical, and performance problems involved in implementing fully recoverable _asynchronous_ OLTP distributed over an internet.

I am interested both in systems with communicating state machines, and stateless systems.

I am particularly interested in papers addressing reliability, lossless recoverability (of operation after a crash), and graceful degradation
(of the system, given one or more components out of action).

Issues involved include whether to use idempotence, how to implement atomicity, etc.

For practical reasons, papers that assume a TCP/IP transport layer are of more interest.

I am also interested in learning about commercial products that attempt
(or succeed?) to solve these problems. I am _not_ interested in bloated
products relying on what are effectively synchronous solutions, or in OSI based solutions.

Please reply via Email, and I will summarise.

Thanks very much for any help,

        -Huw

                         "It's the only way to be sure"

[ Huw Rogers Communications Software Engineer, NEC Corporation, Tokyo, Japan ] [ Email: rogersh_at_ccs.mt.nec.co.jp Fax: +81-3-5476-1005 Tel: +81-3-5476-1096 ] , Received on Tue Dec 14 1993 - 09:29:19 CET