Parallel Databases Michael French, Spencer Steele, Jill Rochelle

12
Parallel Databases Parallel Databases Michael French, Spencer Steele, Jill Rochelle Michael French, Spencer Steele, Jill Rochelle When Parallel Lines Meet by Ken Rudin (BYTE, May 98)

description

Parallel Databases Michael French, Spencer Steele, Jill Rochelle. When Parallel Lines Meet by Ken Rudin (BYTE, May 98). What are Parallel/Scalable Databases?. Parallel/Scalable Databases: Hardware Architecture Multiple Processors Multiple Disk Drives Large Memory Banks - PowerPoint PPT Presentation

Transcript of Parallel Databases Michael French, Spencer Steele, Jill Rochelle

Page 1: Parallel Databases Michael French, Spencer Steele, Jill Rochelle

Parallel DatabasesParallel Databases

Michael French, Spencer Steele, Jill RochelleMichael French, Spencer Steele, Jill Rochelle

When Parallel Lines Meetby Ken Rudin (BYTE, May 98)

Page 2: Parallel Databases Michael French, Spencer Steele, Jill Rochelle

What are Parallel/Scalable What are Parallel/Scalable Databases?Databases?

Parallel/Scalable Databases:Parallel/Scalable Databases: Hardware Architecture Hardware Architecture

Multiple ProcessorsMultiple Processors

Multiple Disk DrivesMultiple Disk Drives

Large Memory BanksLarge Memory Banks Software ArchitectureSoftware Architecture

Capable of processing parallel queriesCapable of processing parallel queries

Data shipping capabilitiesData shipping capabilities

Page 3: Parallel Databases Michael French, Spencer Steele, Jill Rochelle

What makes Parallel Databases What makes Parallel Databases different from previous different from previous

technologies?technologies?

Page 4: Parallel Databases Michael French, Spencer Steele, Jill Rochelle

Previous TechnologyPrevious Technology

HardwareHardwareSingle processorSingle processor

Small Disk CapacitySmall Disk Capacity

Less MemoryLess Memory SoftwareSoftware

Sequential QueriesSequential Queries

No partitioning of queriesNo partitioning of queries

Page 5: Parallel Databases Michael French, Spencer Steele, Jill Rochelle

Parallel Query: Parallel Query:

A Query that partitions information A Query that partitions information to multiple processors and also has to multiple processors and also has the ability to pipeline informationthe ability to pipeline information

Page 6: Parallel Databases Michael French, Spencer Steele, Jill Rochelle

Information PartitioningInformation Partitioning

Divide the information into smaller Divide the information into smaller taskstasks

Can have multiple meanings:Can have multiple meanings:– Distribution of info to multiple CPUsDistribution of info to multiple CPUs– Division of hard drive space to Division of hard drive space to

contain certain parts of the datacontain certain parts of the data

Page 7: Parallel Databases Michael French, Spencer Steele, Jill Rochelle

Information Partitioning 2Information Partitioning 2

Page 8: Parallel Databases Michael French, Spencer Steele, Jill Rochelle

Information PipeliningInformation Pipelining

Allows separate processors to work Allows separate processors to work on separate stages of a queryon separate stages of a query– ScanScan– JoinJoin– SortSort

Concept is akin to assembly line idea Concept is akin to assembly line idea Allows multiple queries to run at the Allows multiple queries to run at the

same timesame time

Page 9: Parallel Databases Michael French, Spencer Steele, Jill Rochelle

Information Pipelining 2Information Pipelining 2

Page 10: Parallel Databases Michael French, Spencer Steele, Jill Rochelle

Sequential Query ExampleSequential Query Example

Two Tables with 20 million rows Two Tables with 20 million rows each run on a uniprocessor machineeach run on a uniprocessor machine– To perform scan, join & sort, query To perform scan, join & sort, query

takes 12 mins.takes 12 mins. Add partitioningAdd partitioning

– Query takes 3 mins.Query takes 3 mins. Add PipeliningAdd Pipelining

– 12 queries can be run in 12 mins.12 queries can be run in 12 mins.

Page 11: Parallel Databases Michael French, Spencer Steele, Jill Rochelle

Parallel KindsParallel Kinds

Share-EverythingShare-Everything– HardwareHardware– SoftwareSoftware

Share-DiskShare-Disk– HardwareHardware– SoftwareSoftware

Share-NothingShare-Nothing– HardwareHardware– SoftwareSoftware

Page 12: Parallel Databases Michael French, Spencer Steele, Jill Rochelle

ConclusionConclusion

ProsPros– Allows you to process more informationAllows you to process more information– Provides for faster processing of queriesProvides for faster processing of queries

ConsCons– Expensive hardware & softwareExpensive hardware & software– Much higher maintenance Much higher maintenance

Is a parallel database right for your Is a parallel database right for your organization?organization?