The recovery
manager should direct the request manager to block new client requests until the connection
log file had been processed. Besides that relation, the recovery manager will also be
interested in view changes. When it receives failure notifications, it will start logging client
activity, and when it receives join notifications, it will start the state transfer protocol.
Conclusion
Automatic replication of existing applications is one of the most interesting challenges in
high availability. Since a huge amount of applications do currently exist that were not designed
to explicitly face failures and recovery, a pluggable highly-available solution has an
enormous role to play. This chapter has detailed a fairly simple software architecture as the
instrument to explain the basic problems and solutions that appear in automatic replication
for Web applications.
We have shown that the most discouraging problem is the set of assumptions that Web applications
should fulfill to benefit from this approach. Replicable actions, or in other words,
self-containment of the full application, could be problematic. However, it can usually be
solved by implementing some extra proxy software modules; but in this case, the solution
cost may rise due to extra development effort.
Pages:
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578