Please enable JavaScript.
Coggle requires JavaScript to display documents.
TECHNOLOGY OF DISTRIBUTED DATABASES - Coggle Diagram
TECHNOLOGY OF DISTRIBUTED DATABASES
What we don't need from Acid
Consistency
: integrity constraints describe only local properties (it is a limit of current DBMS
technology).
Persistency
: each system guarantees persistency to locally stored data (local recovery (log,
checkpoint, dump) mechanisms).
What we need to adjust in Acid
Isolation
We cannot have a module acting as a scheduler operating locally in each node because the serial execution of portions of the transaction do not correspond to the serial execution in the overall system of the same set of transactions. We need a new type of check.
Atomicity
:
we need a coordination for the commit or the abort of the transactions if something
goes wrong, considering the fact that the different fragments DO NOT talk one to another.
Data distribution requires to modify
Query optimization
Concurrency control (the scheduler) to guarantee isolation
Reliability control (distributed commit) to guarantee atomicity
Distributed query optimizer
Query divided into subquery (addressed to a specific dbms)
Query is interpreted as a tree and reorganised in order to produce the same result
Coordinated execution of different programs at different DBMSs
Guarantees global optimization: minimize the response time for the final user
Query optimizer never reaches real optimum, but tries to limit execution costs
Concurrency control