Distributed Databases Questions and Answers

Answer: A

Correct answer: True

A simple snapshot draws on a single table and can often be refreshed by sending only the rows that have changed. A complex snapshot involves joins or aggregation, so determining what has changed may require recomputing it, and a full refresh is frequently cheaper than working out the difference.

Answer: B

Correct answer: False

It is synchronous technology that spends the time, because the transaction cannot commit until every copy has been updated and confirmed. Asynchronous technology returns as soon as the local update is done and propagates afterwards, which is why it is faster and why it tolerates a site being unavailable.

Answer: B

Correct answer: The required data are located in at least one nonlocal site and the distributed DBMS routes requests as necessary.

A transaction is global when it needs data from beyond its own site, and the distributed DBMS then decomposes it and routes the pieces to wherever the data lives. A local transaction is satisfied entirely at one site and needs no such routing - the distinction that decides how much coordination the transaction costs.

Answer: A

Correct answer: True

Failure transparency means a transaction remains atomic across the whole distributed system: either every site commits its part or none does. Achieving it requires a commit protocol such as two-phase commit, in which the sites first agree they can commit and only then do so.

Answer: A

Correct answer: True

A centralised system controls concurrency with one lock manager holding the whole picture. Distributing it means locks held at several sites, messages that may be delayed or lost, and deadlocks whose cycle spans sites and which no single site can see. This is why timestamping, which avoids locks altogether, is attractive in a distributed setting.

Answer: A

Correct answer: True

A full copy at each of two or more sites serves both purposes at once. It distributes the data so that each site can read locally, and it provides fault tolerance, since the failure of one site leaves a complete database elsewhere. Falling storage costs are what have made the approach increasingly attractive.

Answer: A

Correct answer: True

Timestamping assigns each transaction an order in advance and resolves conflicts by that order, so no transaction has to wait for a lock. This removes deadlock entirely, which is a substantial gain in a distributed system where detecting a deadlock spanning several sites is difficult. The cost is that a transaction found to be out of order must be restarted.

Answer: B

Correct answer: False

A network connection is part of the definition. Databases on computers at several locations with no network between them are a decentralised system: nothing can be queried across them and there is no single logical database. Distribution requires the sites to be interconnected.

Answer: A

Correct answer: True

This is the definition. The joining attribute alone is sent to the other site, which matches it against its own rows and returns only those that participate in the join. The saving comes from restricting both what is sent and what comes back, and it is what makes distributed joins practicable over a slow link.

Answer: D

Correct answer: All of the above.

The transaction manager is responsible for atomicity and isolation, so it must log what each transaction does, hold the before and after images needed to undo or redo it, and apply the concurrency control that keeps concurrent transactions from interfering. All three are aspects of the same duty.