Chapter 25 Flashcards
What is data parallelism
One operation performs on multiple partitions simultaneously.
What is query coordinator
It coordinates the results of independent processor
Should query coordinator processing time should less than query server processing time
Yes
What is temporal parallelism (pipe lining)
One complex task divided into smaller parts. The smaller tasks runs in parallel in a way that one sub task output goes into the input of another sub task. Its example is car manufacturing. One car manufactured parts by parts in a factory on an assembly line. And during that manufacturing, another cars manufacturing process starts. The same process goes with complex queries in pipe lining.
What are limitations(issues) of temporal parallelism (pipe lining)
- The relational operator (complex query) can not be divided more than 10 parts. It means the speed-up can not be greater than 10.
- If there is some sub task that needs all inputs to process than it means it is not pipe lining. Because in pipe lining, a sub task takes some inputs and its output is input for another sub task. If one sub task take all input than it is not pipe lining. Aggregation and sorting are its example.
What is full table scan query
All table scans in it.
What is point query
Find a specific value
What is range query
Range is specified in query
What is round robin partitioning technique
It distributes all things equally to processors. There is no bottle neck in it and every processor do its job.
What is hash partitioning technique
In uniform hashing every process has same load. It is best for point query.
What is range partitioning technique
Partitioning on the base of range
What is hot spot or skew
Uneven distribution
What are 2 types of parallelism
- Data parallelism
- Temporal parallelism
