SMP mode
SMP mode with a set of a plurality of processors coupled to a memory. In SMP mode, all processors can access the same physical system memory, which means that SMP systems run only one copy of the operating system.
Thus SMP systems are sometimes also referred to as uniform memory access (UMA) architecture, consistency means that no matter what time, maintaining or processor can share a single value for each data memory.
Obviously, the disadvantage is the limited scalability of SMP, because when the interface reaches saturation in a memory, the processor can not be increased to achieve higher performance.
MPP mode
A distributed memory mode, more processors can be integrated into a system memory.
A pattern memory having a plurality of distributed nodes, each node has its own memory, may be configured as a SMP mode, you may be arranged a non-SMP mode. Individual nodes connected to each other to form a total system.
MPP architecture for hardware developers attractive because they appear relatively easy problem to solve, the development cost is relatively low. Because there is no hardware support issues shared memory or cache coherency, it is relatively easy to achieve a large number of connected processors.
NUMA mode
Also it uses a distributed memory mode, except that all of the processor nodes in the system can access all the physical memory.
However, each processor memory access time required in this node, you may access memory within the time than some remote node spent much less. In other words, memory access time is inconsistent, which is the reason for this pattern is known as "NUMA" of.
In short, NUMA SMP mode while maintaining a single copy of the operating system, features simple application programming model and easy to manage, but also inherited the MPP mode scalability, can effectively expand the size of the system. This is exactly the NUMA advantage.