|
The chances
for parallelism are increased so much because of the availability of computer hardware
for lower prices.
The performance and availability of the system can be enhanced
by using multiple processors which can execute in parallel. This is not same as
the superscalar processor which does the instruction level parallel execution.
Symmetric
Multiprocessors and Clusters are the most common multiple processor organizations.
|
|
Symmetric Multiprocessing
(SMP): This is one of the earliest types of parallel organization. Here a common memory
is shared by the multiple similar processors the same computer which are interconnected
by a bus usually.
We cannot add processors beyond certain limit in order to increase
the performance. So the more scalable organization of systems called clusters is
developed
|
Clusters:
Here
multiple independent computers are grouped in a cooperative manner and makes the
illusion of being one machine. Instead of sharing a common memory, each node has
its own private main memory applications here.
These are generally used for supporting
large work loads and in server applications. But these are complex, consume
more power and occupy more space compared to SMP.
Non Uniform Memory
Access (NUMA): This is a shared memory multiprocessor. Here the access
time from a particular processor to a word in memory changes according to the location
of the memory word
Super Computers Supercomputers are expensive with almost 10 to 15 million
dollars and used for selected applications like research centers and government
scientific agencies. They can perform floating point arithmetic on huge arrays of
numbers with the capacity of doing millions of floating point.
|
