System architecture design professional skills·System engineering and system performance

Table of Contents of Series Articles
1. System engineering
2. System performance

Insert image description here

1. System engineering

1.1. System engineering concept

[System Engineering] is an organizational management technology.
[System Engineering] is a scientific method that analyzes and studies the component elements, organizational structure, information flow, and control mechanism of the system in order to best realize the purpose of the system.
[System Engineering] Starting from the whole, starting from the system concept, in order to achieve the overall optimization .
[System Engineering] Use computers as tools to analyze the structure, elements, information and feedback of the system to achieve optimal planning, optimal design, optimal management and optimal control.
[System Engineering] is a modern scientific decision-making method.

1.2. Systems engineering method

Systems engineering is an organizational management technology that uses systems methods to plan, research, design, manufacture, test and use systems. It is a technology that people use scientific methods to solve complex problems .

The characteristics of the systems engineering method : integrity, comprehensiveness, coordination, scientificity and practicality .

Systems engineering methods are divided into : Hall's three-dimensional structure, Checkland method, concurrent engineering, comprehensive integration method, and WSR system method .

Insert image description here

1.2.1 Hall’s three-dimensional structure

Hall's three-dimensional structure is a systems engineering methodology proposed by American systems engineering expert A.D. Hall and others in 1969. It forms athree-dimensional space structure composed of time dimension, logical dimension, and knowledge dimension .

Time dimension: divided into 7 time stages of planning, formulating plans, development, production, installation, operation, and update; logical
dimension: including 7 logical steps of clarifying problems, determining goals, system synthesis, system analysis, optimization, decision-making, and implementation ;
Knowledge dimension: includes knowledge and skills in engineering, medicine, architecture, business, law, management, social sciences, art, etc.

Insert image description here

1.2.2 Checkland method

The core of Checkland's method is not "optimization" but "comparison" and "exploration" . The work process is divided into seven steps: understanding the problem, basic definition, establishing conceptual model, comparison and exploration, selection, design and implementation, evaluation and feedback .

1.2.3 Concurrent Engineering

The concurrent engineering method is a systematic method and comprehensive technology for parallel and integrated processing ofproducts and related processes (including manufacturing processes and support processes) . The goal is to improve quality, reduce costs, shorten product development cycles and product time to market.

1.2.4 Comprehensive integration method

Qian Xuesen and others proposed that systems can be divided into two categories : simple systems and giant systems based on the nature of the system. The general basic principles of open complex giant systems are: holism, interconnectedness, orderliness, and dynamics . The main properties are openness, complexity, evolution and emergence, hierarchy, and massiveness.

1.2.5 WSR System Method

WSR System Method is the abbreviation of Physics-Philosophy-Humanity methodology. The speculative thought of traditional Chinese philosophy is the comprehensive unity of multiple methods and belongs to the Eastern systematic thought that comprehensively integrates qualitative and quantitative analysis. The general work process can be understood as seven steps: understanding intentions, setting goals, investigation and analysis, constructing strategies, selecting plans, coordinating relationships, and realizing ideas.

1.3 Life cycle of systems engineering

The purpose of defining the systems engineering life cycle is to establish a framework that meets the needs of stakeholders in an orderly and efficient manner.

The life cycle stages of system engineering include exploratory research, concept stage, development stage, production stage, use stage, support stage and decommissioning stage.

Life cycle methods include: plan-driven approach, progressive iterative development, lean development and agile development.

1.4 Model-based systems engineering

Model-Based Systems Engineering (MBSE) , MBSE is a formal application of modeling methods so that modeling supports activities such as system requirements, analysis, design, verification and validation, and continues through all life cycle stages. .

Products include :
in the requirements analysis stage , demand diagrams, use case diagrams and package diagrams are produced;
in the function analysis and allocation stage , sequence diagrams, activity diagrams and state machine diagrams are produced;
in the design synthesis stage , module definition diagrams, internal block diagrams and Parametric diagrams, etc.;

The three pillars of systems engineering :
modeling languages, modeling tools and modeling ideas.

2. System performance

2.1 System performance indicators (evaluation)

System performance evaluation indicators are an integration of software and hardware performance indicators . in:

(1) The main performance indicators for evaluating computers include clock frequency (main frequency), operation speed, operation accuracy, data processing rate (Processing Data Rate, PDR), throughput rate, etc.
(2) The main performance indicators for evaluating routers include device throughput, port throughput, full-duplex line-speed accurate transmission capability, routing table capability, backplane capability, packet loss rate, delay, delay jitter, protocol support, etc. The performance indicators used to evaluate switches include port speed, backplane throughput, buffer size, MAC address table size, etc.
(3) The evaluation network performance indicators include equipment performance indicators, network-level performance indicators, application-level performance indicators, user-level performance indicators and throughput .
(4) The performance indicators for evaluating operating systems include system context switching, system response time, system throughput (volume), system resource utilization, reliability and portability .
(5) The main performance indicators for measuring the database management system include maximum concurrent transaction processing capability, load balancing capability, maximum number of connections, etc.
(6) The main performance indicators for evaluating Web servers include the maximum number of concurrent connections, response delay and throughput .

Insert image description here

2.2 Performance calculation

Performance Calculation
Insert image description here
Clock frequency and CPU clock cycle:

The main frequency is also called the CPU clock frequency, and the clock cycle is the reciprocal of the clock frequency.
If the main frequency is 1GHz, it means that there are 1G clock cycles in 1 second, and each clock cycle is 1ns.
Main frequency = FSB * Multiplier
Bus cycle:

That is, the time it takes to access a memory or IO port. Instruction
cycle:

The time to fetch and execute an instruction is composed of several machine cycles.
The average number of execution cycles:

CPI represents the clock cycles required to execute each computer instruction, sometimes referred to as the average number of cycles of instructions.
CPI average number of execution cycles = number of clock cycles required to execute the program/number of instructions executed.
MIPS average instruction execution speed:

The number of millions of machine language instructions processed per second is mainly used to measure scalar machine performance
MFLOPS:

Millions of floating point operations per second cannot reflect the overall situation, but only the floating point operations.
Mainly used to measure vector machine performance

Average number of instruction clocks = (number of instruction entries × number of instruction clocks)/number of total instruction entries
Average instruction operation (execution) speed = 1/average number of instruction clocks × clock frequency, note, unit MIPS

2.3 Performance design

2.3.1 Performance Tuning

Performance tuning consists of finding and eliminating bottlenecks.

For database systems :
Performance adjustment mainly includes CPU/memory usage, optimized database design, optimized database management and process/thread status, hard disk I/O and remaining space, log file size, etc.

For application systems :
Performance adjustment mainly includes the availability of the application system, response time, number of concurrent users, and system resource usage of specific applications, etc.

2.3.2 Amdahl solution

Amdahl's Law: The degree of change in system performance obtained by using a faster execution method for a certain component in a computer system depends on the proportion of the total execution time occupied by this method.

Speedup ratio = time to complete the entire task when using enhanced components / time to complete the entire task when not using enhanced components

New execution time = original execution time X [ (1 - enhancement ratio) + enhancement ratio / enhancement speedup ratio ]

Total speedup ratio = original execution time / new execution time = 1 / [ (1 - enhancement ratio) + enhancement ratio / enhancement speedup ratio ]

2.4 Performance evaluation

(1) Benchmark testing program

In most cases, to test the performance of a new system, users must rely on evaluation programs to evaluate the performance of the machine.

The definition of benchmark program (Benchmark) :
the core program that is used the most and most frequently in an application as a standard program for evaluating computer performance.

Four evaluation programs are listed below, in order of decreasing accuracy :
real program, core program, small benchmark program, and synthetic benchmark program .

The benchmark testing programs include :
integer testing program Dhrystone, floating point testing program Linpack, Whetsone benchmark testing program, SPEC benchmark testing program and TPC benchmark program .

(2) Performance evaluation of Web server

Web server performance evaluation methods include :
benchmark performance testing, stress testing and reliability testing .

(3) System monitoring

System monitoring methods usually include three methods :
system built-in commands, system log review, and visualization technology .