Introduction to MPP, SMP, NUMA concepts

First, Mpp,smp,numa Concept Introduction

1.1. MPP Architecture Introduction

The MPP (massively Parallel processing), massively parallel processing system, is made up of a number of loosely coupled processing units, and it is important to note that this refers to the processing unit rather than the processor. The CPU in each unit has its own private resources, such as bus, memory, hard disk, etc. Each unit has an operating system and an instance replica of the management database . The biggest feature of this structure is the non-sharing of resources.

1.2. SMP Architecture Introduction

SMP (symmetric Multi processing), symmetric multi-processing systems have many tightly coupled multiprocessor, in which all CPUs share all resources, such as bus, memory and I/o system, or only one copy of the operating system or the management database, one of the biggest features of this system is the sharing of all resources .  

1.3. NUMA Architecture

NUMAServerThe basic feature of this is that it has multiple CPU modules, each CPU module consists of multiple CPUs ( such as 4 ) , and has independent local memory, I/O slots and so on. Connections and information interactions can be made between their nodes through interconnected modules , such as Crossbar Switch .

Differences between the two or three architectures

2.1 ,NUMA ,MPP ,SMP the difference between performance

numa The node interconnect mechanism of the is implemented within the same physical server, when a cpu When remote memory needs to be accessed, it must wait, which is also numa Server cannot implement cpu The performance linearly expands when increases.

the node interconnection mechanism of MPP is implemented by I/O on different SMP servers, and each node accesses only local memory and storage, and the information interaction between nodes is carried out in parallel with the processing of the node itself. Therefore, the performance of MPP can be linearly extended when adding nodes.  

All of the CPU Resources of SMP are shared, so linear scaling is fully implemented.

2.2 ,NUMA,MPP,SMP, the difference between extensions

NUMA can theoretically be infinitely extensible, and the technology is now more mature to support hundreds of CPUs to scale. such as HP 's SUPERDOME.

mpp The can theoretically be infinitely extended, and now the technology is more mature to support nodes, thousands of cpu is extended.

SMP poor scalability, currently 2 to 4 < Span style= "font-family: ' Times New Roman ';" >cpu is the best utilization, but IBM Span lang= "It" xml:lang= "it" >book technology, capable of cpu extended to 8 .

MPP is composed of multiple SMP , multiple SMP servers connected through a certain node internetwork, work together to accomplish the same task .

2.3 , MPP and the SMP , NUMA the difference between apps

MPP Advantages of:

mpp the system does not share resources, so for it, the resources are more than SMP more, when the transaction to be processed reaches a certain scale, is more efficient than smp good. Because mpp The system is mpp system can give full play to the advantages of resources to achieve high efficiency. In other words: The operation is not related to each other, the processing unit need to communicate less, that the use of mpp system is better. Therefore, the mpp system shows an advantage in decision support and data mining.

SMP Advantages of:

The MPP system is less efficient than SMP because it wants to transfer information between different processing units . When the communication time is high, the MPP system can give full play to the advantages of resources. so the currently used OTLP program, the user accesses a central database, if using the SMP system structure, it is more efficient than using MPP structures are much faster.

NUMA Benefits of Architecture:

NUMA Architecture, it can integrate many CPUswithin a physical server, enabling the system to have high transaction processing power, as far as memory access Shiyan longer than local memory access, it is necessary to minimize different CPU Data interactions between modules. Obviously, theNUMA architecture is more suitable for the OLTP transaction processing environment, and when used in the Data Warehouse environment, the CPU utilization will be greatly reduced because of the large amount of complex data processing which inevitably leads to a lot of interaction.

Introduction to MPP, SMP, NUMA concepts