IT Infrastructure Basics

While learning about IT infrastructure, I found a very well-organized article and studied while reorganizing it!

References: futurecreator.github.io

1. What is IT Infrastructure?

What is IT Infrastructure?

• The foundation of a system including hardware, OS, middleware, network, etc. needed to run an application

• Related to non-functional requirements

Functional Requirements and Non-Functional Requirements

1. Functional requirement

   • What functions the system has and what it can do

2. Non-functional requirement

   • Requirements such as system performance, stability, security, etc.

Components of Infrastructure

• Hardware (HW)

  • Server hardware, storage for data, power supplies, etc.

    • In a broader sense, this also includes data center facilities where hardware is installed

• Network

  • Tools that connect servers so users can access them remotely

    • ex)

      • Network equipment such as routers, switches, firewalls

      • Cable wiring connecting them

      • Access points (AP) needed for users to connect wirelessly from terminals

• Operating System (OS)

  • Basic software for controlling hardware and network equipment

  • Manages resources and processes

    • Client OS

      • Focuses on making it easy for users to use

      • ex) Windows, macOS

    • Server OS

      • Focuses on running systems quickly and stably

      • ex) Linux, Unix, Windows Server

• Middleware

  • Software that provides functionality to make the server perform a specific role on the server

2. On-premises vs Cloud

2-1. On-premises

• The traditional approach of placing servers in a data center or server room and managing them directly

• Required purchasing, installing, and managing servers, network equipment, OS, storage, various solutions, etc.

Disadvantages of On-premises

• Equipment is expensive, so initial investment costs are high

• Difficult to estimate usage, so once built, maintenance costs remain the same even with low usage

2-2. Public Cloud

• A system provided as a service to an unspecified majority through the internet

  • What does service form mean?

    • Users select desired options and pay for what they use

    • ex) IaaS (select VMs or storage with desired specs, pay based on usage time or data volume), PaaS, SaaS

• Cloud Providers like AWS, Microsoft Azure, GCP own data centers and infrastructure

2-3. Private Cloud

• A form of Public Cloud with limited users

• ex) Internal corporate services - Good security, easy to add proprietary features or services

2-4. Pros of Cloud

• Cloud is advantageous for systems with high traffic variability

  • Why?

    • Because it is not easy to predict traffic for external services!

    • Estimating server specs or network bandwidth based on traffic volume is called Sizing, which is quite a difficult task

      • Cloud systems have Auto Scaling that automatically scales based on traffic fluctuation, making it more advantageous than On-premises!

• Even if a data system goes down due to natural disasters, the system can be operated from elsewhere

  • How?

    • Because cloud data centers are spread around the world!

• Cloud is also advantageous for systems that need to provide services quickly or PoC (Proof of Concept)

  • Also advantageous for startups / individual developers with low initial investment!

2-5. Pros of On-premises

• Both On-premises and Cloud guarantee availability, but there is a difference in concept

  • On-premises

    • Aims for servers not to go down

  • Cloud

    • In a distributed environment with many instances, if an instance goes down, another instance quickly replaces it

      • In other words, using Cloud does not guarantee availability; you must design it to increase availability yourself

      • Therefore, On-premises is advantageous for systems that must never be interrupted even briefly or need availability beyond what cloud providers guarantee...

        • That being said, I can't quite agree with this part..!

          • Taking Amazon RDS as an example, there are many methods that come to mind right now, such as using multi-AZ deployments for failover, provisioning and maintaining synchronous standby replicas in other availability zones, etc., to increase availability!!

          • I wonder if On-premises can guarantee availability beyond what Cloud can offer! I should look into it

• On-premises is advantageous for data with high confidentiality

  • Of course, the security provided by Cloud Providers may be better than your own, but On-premises is advantageous when you need to know the exact physical storage location

  • Also, if using multi-cloud, security standards are difficult to establish because each Cloud provider has different security policies...

    • I have joined a Multicloud project in the new open source contributhon, and I hope to find out if this part is really difficult during the contributhon! I should look into it

2-5. Hybrid Cloud

• Since both On-premises and Cloud have their pros and cons, both are sometimes used together

  • Using both according to the characteristics of each system!

• Since cloud providers also have different strengths and weaknesses, multiple clouds are sometimes used together

• To make good decisions about this, you need to know the characteristics of each well, and the criteria for selection must be clear!

3. Infrastructure-Related Concepts

3-1. Hardware

• Hardware and Network are the most low-level elements of Infrastructure

• In On-premise systems, it consists of multiple server machines

• In Cloud, you select hardware performance of instances as needed

3-2. CPU

• CPU performance is influenced by Core and Cache

  • The more Cores, the more concurrent computations can be processed

  • Cache, which alleviates the processing speed gap with memory, performs better with larger size

• GPU

  • A processor specialized in graphics processing

  • While CPU consists of a few cores optimized for serial processing,

    • GPU consists of many small cores optimized for parallel processing

  • In fields that require high-speed processing of large amounts of data, such as deep learning or numerical analysis, GPU Computing is used to boost processing performance by using CPU and GPU together

    • This method offloads computation-heavy parts to the GPU and processes only the remaining code on the CPU! Interesting..

3-3. Memory

• Main memory performs better with larger data capacity and faster transfer speed

• For server use, memory with low power consumption and built-in error handling is typically selected

3-4. Data Storage

• A device that stores data

• Since storage is usually the slowest component, the storage capacity and read/write speed often affect the overall system speed

• Consists of hard disks, SSD (Solid-State Drive), etc.

Data Management

• The most important thing in IT can be said to be data

• Since this data must not be lost, most systems are configured with redundancy or multiple redundancy for high availability (HA, the ability to operate continuously over a long period)

• ex)

  • Configuring redundancy by distributing AWS RDS across different Availability Zones (AZ) within the same Region

    • AWS calls this Multi-AZ Deployment

    • Even within the same region, data is distributed so if the Master has a problem, the Slave recovers and maintains data

3-5. Firewall

• A firewall controls communication between the internal network and external network for security, blocking unnecessary communication

• ex)

  • Packet filtering type

    • Filters passing packets based on port numbers and IP addresses

  • Application gateway type

    • Controls communication with the outside at the application protocol level, commonly called a proxy server

    • To inspect information contained in sessions, it terminates the existing session and creates a new one

    • Slower than packet filtering type but can perform more inspections

4. Network

4-1. Network Addresses

Networks use network addresses to identify various equipment

1. MAC Address (Physical Address / Ethernet Address)

: A physically assigned 48-bit address

• The first 24 bits identify the network component manufacturer

• The last 24 bits are assigned uniquely by each manufacturer

• Expressed in hexadecimal, separated by 2 bytes

2. IP Address

: A number assigned to equipment connected to a network such as the internet or intranet

• The commonly used IPv4 is a 32-bit address divided into 4 groups of 8 bits each

  • ex) 192.168.1.1

  • Each position can represent 0 to 255

• IPv4 can only connect up to 2^32 (approximately 4.2 billion) devices on a single network, raising concerns about IP address exhaustion on the internet

  • That is why IPv6 uses 128-bit IP addresses!!

• Internal networks use private addresses with arbitrary address assignments and use NAT devices to convert to global addresses at the boundary with the internet

4-2. OSI Model

• When communicating, 1) rules for how to exchange messages and 2) what language to use are needed

  • These conventions are called communication protocols

• The OSI (Open Systems Interconnection) Model is a model that divides computer communication functions into a layered structure, created by the International Organization for Standardization (ISO)

• Using the OSI Model, you can visually understand what happens in a specific networking system using layers

• The OSI Model consists of a total of 7 layers

  • Data -> when going out to the network, from the top layer

  • Network -> when receiving data, from the bottom layer

OSI Model vs TCP/IP Model

1. Physical Layer

• The layer where transmission cables are directly connected, functioning to transmit through cables

  • The function of sending and receiving data!

• Defines voltage & current values, and specifies the physical and electrical characteristics of communication equipment such as cable and connector shapes

  • ex) Twisted pair cables (STP/UTP) used for LAN cables, Ethernet standards like 100BASE-T, or IEEE802.11 series wireless communication

2. Data Link Layer

• Defines communication between two adjacent systems (Nodes) within the same network

  • Responsible for node-to-node delivery!

• Verifies that the physical layer is functioning properly

• Receives data packets from the network layer and adds MAC addresses and various control information

  • The data unit with additional information is called a frame, which is transmitted through the physical layer

3. Network Layer

• Defines rules for communication between different networks

• Has a routing function that efficiently processes paths to specific servers

• While the data link layer is based on MAC addresses, the network layer is based on IP addresses

• Responsible for safely delivering data from source to destination

  • For this purpose, it has:

    • Flow control function that controls packet flow when there is heavy packet traffic

    • Error control function that detects packets lost during transmission and requests retransmission

• Representative equipment includes routers and L3 switches

  • Such equipment manages a routing table that stores information about where to forward packets

    • Static route that determines routes based on the routing table

    • Dynamic route set by routing protocols

  • An L3 switch is equipment that processes the same functions as a router in hardware!

4. Transport Layer

• The layer that controls data transmission

• Handles the volume, speed, and destination of data to be sent

• The session layer's messages are divided into segments, with sequence numbers recorded for each segment, sent to the network layer, and reassembled on the receiving side

  • This method specifies transmission error detection and retransmission

• Representative protocols include TCP and UDP

5. Session Layer

• Responsible for maintaining and releasing connections between applications

• Defines connection establishment timing and data transmission timing, etc.

6. Presentation Layer

• Responsible for converting data so applications can understand it

• Handles conversion of data storage formats, compression, character encoding, and also processes encryption and decryption functions for secure data transmission

7. Application Layer

• The top layer, referring to applications directly used by users such as web browsers or Outlook

5. Linux

Now that we've covered hardware and networking, let's look at the OS

What is Linux?

• Linux is a Unix-compatible server OS developed by Linus Torvalds

• According to the Linux Foundation:

• 90% of public cloud workloads

• 82% of the world's smartphones

• 62% of embedded devices

• 99% of the supercomputer market runs on Linux

• Linux is open source, built with the participation of various companies and individuals

  • It is said to be the open source project with the most contributors in computer history!

Linux Kernel

• The Kernel is the core part of the OS

• Roles of the Linux Kernel:

  • Memory management

  • File system

  • Process management

  • Device control

• Android is also built on the Linux Kernel

1. Device Management

• The Linux Kernel controls hardware such as CPU, memory, disk, and IO using software called device drivers

2. Process Management

• Linux reads the contents written in program files and processes them in memory; the executed program is called a process

• Processes are managed by assigning PID (Process ID) and efficiently allocating necessary resources to each process

3. Shell

• Users can send commands to the kernel through a Command Line Interface called Shell

• A collection of commands to be executed in the Shell is called a Shell Script

  Shell Name	Features
  bash	- Supports command history, directory stack, and auto-completion for commands and file names - Comes standard in most Linux systems and macOS (OS X)
  csh	- A Shell similar to the C language - Mainly used in BSD-based OS
  tsch	An improved version of csh that supports auto-completion for commands and file names
  zsh	A Shell compatible with bash that operates at high speed

4. File System

• A file system is a scheme that names files and indicates where to store them

  • A system for managing files

• The Linux Kernel uses a Virtual File System (VFS)

• It allows users to use data as if they were regular files regardless of where they are stored (hard disk, memory, network storage, etc.)!

• The Virtual File System (VFS) treats each device as a file

• ext2

  • A file system widely used in the Linux operating system

  • Called ext2 because it extended the initial ext file system

• ext3

  • A file system mainly used in Linux

    • Available from Linux Kernel 2.4.16!

• ext4

  • The successor file system to ext3

  • Storage supports up to 1EiB

  • Supports extent file writing that prevents file fragmentation

• tmpfs

  • A device for temporary files in Unix-based OS

  • Can be stored in memory

  • Often mounted as /tmp, and since it is stored in memory, all files are lost when the server restarts!

    • That's right..... they all disappear..

• UnionFS

  • A file system that can layer multiple directories and treat them as a single directory

• NFS

  • A distributed file system and protocol used in Unix

5. Directory Structure

• Linux's directory listing is standardized by a specification called FHS (Filesystem Hierarchy Standard)

  • Most major distributions compose their directories based on FHS

Directory	Description
/	Root directory
/bin	Directory storing basic commands like ls, cp
/boot	Directory storing files needed for OS booting, such as the Linux Kernel
/dev	Directory storing device files for hard disks, keyboards, etc.
/etc	Directory storing configuration files for OS or applications
/home	- Home directory for regular users - Root user uses /root as home directory!
/proc	- Directory storing information about kernel and processes - Numbered folders under /proc represent Process IDs!
/sbin	Directory storing system administration mounts
/tmp	- Directory for temporarily used files - Deleted when the server restarts
/usr	Directory storing various programs and kernel source
/var	Directory storing files that change with system operation

6. Security Features

• Account-based permission settings

  • Linux can set permissions for user accounts

  • There is a root user who manages the entire system and other regular users

    • There are also system accounts for running middleware daemons

  • Accounts can be grouped!

  • Based on accounts and groups, access permissions for files or directories can be set

• Network filtering

  • Linux was originally built as an OS intended for multiple users on a network, so it has many network-related features

  • iptables is a feature built into Linux that can configure packet filtering and NAT

• SELinux (Security-Enhanced Linux)

  • SELinux is provided by the U.S. National Security Agency, adding mandatory access control features to the Linux Kernel

  • In Linux, the root user can access everything regardless of permissions, so if the root account is compromised, it can have critical impact on the system

    • SELinux prevents power concentration on root with TE (Type Enforcement) that restricts access per process and RBAC (Role-based Access Control) that controls all users including root

      • Interesting!!

7. Linux Distribution

• Usually, Linux is distributed as packages called distributions that include various commands, libraries, and applications on top of the kernel

• There are various distributions because different people want different programs and Linux is open source, allowing individuals or companies to modify and use it

• Distribution Types

  • Debian-based

    • Debian

      : Linux developed by the GNU/Linux community

    • KNOPPIX

      : Linux that can be used with CD booting

    • Ubuntu

      : Linux providing a rich desktop environment Ubuntu is the best

  • Red Hat-based

    • Fedora

      : Linux from the Fedora Project community supported by Red Hat

    • Red Hat Enterprise Linux (RHEL)

      : Commercial Linux provided by Red Hat

    • CentOS

      : Linux aiming for complete compatibility with RHEL

    • Vine Linux

      : Linux developed in Japan

  • Slackware-based

    • openSUSE

      : Linux developed by a community supported by Novell

    • SUSE Linux Enterprise

      : Stabilized commercial Linux based on openSUSE

  • Other Distributions

    • Arch Linux

      : Linux that uses Pacman as its package management system

    • Gentoo Linux

      : Linux that uses Portage as its package management system

6. Middleware

• Middleware refers to various software that sits between the OS and business-processing applications

  • ex) Web Server, DBMS, System monitoring, etc.

6-1. Web Server

• A Web Server is a server that receives HTTP requests from clients and returns Web Content as responses or invokes server-side applications

Name	Description
Apache HTTP Server	A widely used traditional open source web server
Internet Information Services	- A web server provided by Microsoft - Included with OS products like Windows Server series
Nginx	Has low memory consumption

6-2. DBMS

• A Database Management System (DBMS) is middleware that manages databases

• It includes basic functions such as CRUD (Create, Read, Update, Delete) and many features including Transaction processing

• There are various types of DBMS

• Since syntax varies significantly between vendors, there is a standard called ANSI SQL

• Since each DBMS varies greatly in supported features, performance, and price, choose according to the required purpose!

RDBMS

Name	Description
Oracle Database	- A commercial RDBMS provided by Oracle - A database widely used by enterprises, with the #1 global DB market share - Provides many features at a correspondingly high price
MySQL	- An open source relational (Relational) DBMS provided by Oracle - The most widely used open source RDBMS - Sun acquired the maker MySQL AB, and then Oracle acquired Sun, making Oracle the owner - Split into a free community version and a paid commercial version - Later, the open source community created MariaDB based on MySQL
Microsoft SQL Server	- A commercial RDBMS provided by Microsoft - Specialized for Windows
PostgreSQL	An open source RDBMS with the 4th largest global market share after Oracle, MySQL, and SQL Server

NoSQL

• NoSQL refers to DBMS that does not use only SQL

• Instead of tables, data is stored in other formats

• Rather than one being better than the other compared to RDB, it is important to use whichever fits the purpose!

Type	Description	Examples
Key-value	- The simplest form of NoSQL - Simple, so it is fast and easy to learn - Queries based on value content are not possible, so processing at the application level is necessary	Redis Amazon DynamoDB Memcached
Document	- Similar to Key-Value but stores hierarchical documents instead of simple values - Queries can be used but differ from standard SQL	MongoDB Couchbase
Wide column stores	- Uses tables, rows, and columns but unlike RDB, column names and formats can differ even within the same row - A 2-dimensional key-value format	Cassandra HBase
Graph	- Stores data as continuous nodes, edges, and properties like a graph - Suitable for data relationship-focused use cases like SNS, recommendation engines, pattern recognition	Neo4j

6-3. System Monitoring

• System operation requires continuous monitoring of various states

• Monitoring is performed at various levels including Network, Server, Cloud, Application, Service, and Transaction to check for anomalies and analyze causes

Name	Description
Zabbix	- An open source monitoring tool developed by Zabbix SIA - Can monitor the status of various servers
Datadog	- A server monitoring SaaS developed by Datadog - Can monitor from a web browser without needing a separate server - Easy monitoring even in multi-cloud environments
Mackerel	- A server monitoring SaaS developed by Hatena - Useful for cloud server monitoring

7. Infrastructure Configuration Management

Infrastructure configuration management refers to the work of managing configuration information for hardware, network, OS, middleware, applications, etc. that make up the infrastructure, and maintaining them in an appropriate state

Immutable Infrastructure

• In On-premise environments, building an infrastructure environment is a big task, and once built, it is used for a very long time while organizing change history

  • But with Cloud, since it is a virtual environment, you can build when needed and discard immediately when not needed!

    • In other words, when a service is updated, instead of changing the existing operating environment, a new image is created and deployed

      • This is called Immutable Infrastructure, meaning it does not change

immutable_infrastructure

• Immutable Infra allows easy server creation from a single image, and since only that image needs to be managed, management is also easy!

• Since the environment itself is deployed, testing in an identical environment is also easy

Still studying...