Distributed Database Question Bank with Answers
1. Is pipelined parallelism part of interoperation parallelism or intra-operation parallelism?
Pipelining is a basic form of interoperation parallelism since a single
query’s different operations are executed in parallel: there are many
processors that are each performing one step in a multi-step process.
(The other form of interoperation parallelism is independent
parallelism where operations of a query that don’t depend on one another are
run in parallel.)
2. List a few relational operators that could be run in parallel.
Most common are:
Scan (combination of SQL’s select and
where-clauses, known as project and select in relational terms),
Sort,
Join, and
Aggregate operators (max, min, sum, count).
3. Why are parallel databases usually homogenous?
Because having a different database schema would make it difficult to
processes the same queries over different processors.
Parallel database architecture helps in improving the speed of query /
transaction execution. Homogenous means identical software, and identical
database design.
If we execute a query on multiple machines, then each must be identical
to improve the performance. If database schema is different for different machines,
the conversion of divided work and compilation of results will be overhead.
4. In what way does shared nothing architecture resemble a distributed database?
Each node has its own processor, disk(s), memory and runs its own
database management software and operating system. Communication with other
nodes is made through the interconnection network.
Most solutions for distributed databases such as distributed
transactions (2PC) and fragmentation can be used in a shared-nothing
architecture.
5. Why is inter-query parallelism in shared memory architecture very straightforward to implement? What problem can occur if inter-query parallelism is implemented in shared disk architecture?
With inter-query
parallelism, the different queries or transactions execute in parallel. Hence, traditional DBMS tricks will work since
the DBMS already has the capacity for multi-threading.
In shared disk
architecture, there is the problem of cache coherence since each processor can
access any part of the database. The system must ensure that each processor has
the latest version of the data in its memory.
******************
Download
Related links:
- Go to Distributed Database Quiz page
- Go to Database Management Systems Quiz page
- Go to Distributed Database Home page
No comments:
Post a Comment