English |  Español |  Français |  Italiano |  Português |  Русский |  Shqip

Big Data Dictionary

Parallel Databases

Parallel machines are becoming quite common and a ordable as the prices of microprocessors, memory and disks have dropped sharply. In parallel database systems, data can be partitioned across multiple disks for parallel I/O where individual relational operations (e.g., sort, join, aggregation) can be executed in parallel. In particular, data can be partitioned and each processor can work independently on its own partition. Hence, di fferent queries can be run in parallel with each other. Large-scale parallel database systems have been increasingly used for: storing large volumes of data, processing time-consuming decision-support queries and providing high throughput for transaction processing. Example of parallel database systems include Asterdata, Greenplum, IBM Netezza, Vertica, Teradata and Infobright.

The above figure illustrates four different architectures of parallel database systems which are described as follows:

a) ˆShared memory: where parallel processors share a common memory. This architecture relies on extremely efficient communication between processors. The main limitation of this approach is that it is not scalable beyond 32 or 64 processors.

b) Shared disk: where parallel processors share a common disk. This architecture can scale to a somewhat larger number of processors but communication between processors is slower and the bottleneck becomes the interconnection to the disk subsystem.

C) Shared nothing: where processors share neither a common memory nor common disk. In this architecture, data accessed from local disks (and local memory accesses) do not pass through interconnection network, thereby minimizing the interference of resource sharing. Shared-nothing multiprocessors can be scaled up to thousands of processors without interference. The main drawback of this architecture is the cost of communication and non-local disk access; sending data involves software interaction at both ends.

d) Hierarchical: Combines characteristics of shared-memory, shared-disk, and shared-nothing architectures.

There has been error in communication with Booktype server. Not sure right now where is the problem.

You should refresh this page.