Design of Parallel Databases – This article is about the Design of Parallel Databases and describes the various types of Shared Architecture in Parallel Databases.
Table of Contents
Parallel DBMS breaks the query operation into different sub-queries to run on different computers to maintain the high-speed transfer rate. Parallel databases use some resources, such as multiple CPUs, disks, memory etc., to establish the efficiency between the computers and return the result as expected from a single computer.
Parallel Database Architecture
- Machines are physically close to each other, e.g., same server room.
- Machines are connected to high-speed Internet connections.
- Communication cost is assumed to be small.
- Can use shared memory, shared disk, or shared-nothing architecture.
Shared Memory Architecture
- Every processor has its disk
- Single memory space is allotted to every processor
- Reading and writing to far memory can be slightly more expensive
- Every processor can have its memory and cache as well
In Shared Memory Architecture, every processor is connected to its disk, which is shared via an interconnection network. A single or global main memory is allotted to all processors. The fast interconnection network allows processors to access memory in parallel.
Shared Disk Architecture
- Every processor has its memory.
- All machines can access all disks in the system
- The number of disks does not necessarily match with the number of processors
In this architecture, multiple processors are connected with interconnection networks. In this, each processor has its own memory and all of them have access to the same disk.
Shared Nothing Architecture
- Most common architecture nowadays, which is used by most companies
- Every machine has its memory and disk; machines can not switch memory and disk to share with each other
- Communication between a machine and shared memory & disk is done with high network and switches
In Shared Nothing Architecture, every processor has its memory & disk, which are not shared with any other machine. Communication between shared memory & disk is established with high-speed networks and switches. It is very common nowadays to use the Shared Nothing Architecture because it is much faster than other shared architectures; therefore, it is used by most companies to manage parallelism.
