1.17 20:13:29 Cleaned up for doxygen.
2004/11/06 mbalazin Bug fix... sigh... fixed case when a replica should stay stuck in unstable state while it receives corrections from upstream.
2004/11/04 mbalazin Bug fixes for running with very long failures.
2004/11/03 mbalazin Adaptive reconciliation algorithm... finally.
2004/10/27 mbalazin - Added support for blocking style failure recovery
- Changed trimming of undo buffers
- other small changes.
2004/10/27 mbalazin Changed protocol between replicas so the one reconciling has time to
finish before getting clients back
2004/10/25 mbalazin just changes in what gets logged
2004/10/23 mbalazin Modified checkpoint() method to re-use BoxPackage.
Making join and the consistency manager log some measurements
2004/10/22 mbalazin - SUnion and SOutput now both inherit from an SControlQBox that holds all
the common stuff for boxes with an output control stream
- ConsistencyMngr now listens for inputs from both types of control streams.
- SOutput signals when reconciliation is completed (instead of SUnion)
- Also bug fix when suspending data path.. cannot suspend it completely
or control messages don't get through
2004/10/22 mbalazin Starting to have Borealis nodes collaborate to handle clients while reconciling their states.
2004/10/19 mbalazin Mostly cleanup in my stuff after all the changes but I also wrapped TupleType into a structure. So we now have to use TupleType::INSERTION rather than INSERTION alone. We can also print tuple types with "<<" and they will print with their string equivalents. We can still compare them with == and !=.
2004/10/19 mbalazin Changed all of my fault-tolerance stuff to use the new tuples headers.
2004/10/12 mbalazin Made fault-tolerance boxes support parameters that define the configuration to run (either checkpoint/redo, undo/redo, only logging, etc.)
2004/10/05 mbalazin - Checkpointing state of Join operators.
- Doing checkpoint/redo to reconcile state of joins when necessary.
- Added a void unpackQueue(const QueueState& packed_queue); method to QBox.
2004/09/28 mbalazin To avoid modifying the Join operator code, I added a user-defined join operator that supports state reconciliation after failures.
I noticed that our Join operator doesn't manage its windows quite correctly but I don't want to touch that at the moment. The problem appears when the window is small and tuples arrive in large batches rather than roughly simultaneously.
I only added a "clear()" method to our current join operator. The method clears the operator state.
Minor improvements in ConsistencyMngr and SUnionQBox.
2004/09/28 mbalazin reconciling with undo/redo
2004/09/27 mbalazin Added a ConsistencyMngr class to the QueryProcessor. This class manages the consistency of the state of the query network in case of failures and network partitions. If you don't care about any of this, you should not see any difference. Nothing happens unless a query network uses an sunion operator.
