Server Components
This is a guide on server components.
Check out Audible on Amazon and listen to the newest books!
The Benefits of Server Usage
Servers play a critical and indispensable role in modern computing environments, serving as the backbone of organizational information technology infrastructure. When discussing servers in contemporary contexts, it is essential to recognize that the term encompasses both physical servers, which are tangible hardware devices, and virtual servers, which are software-based representations of physical servers that may operate either on-premises within an organization's own facilities or in cloud-based environments managed by third-party providers. The importance of servers cannot be overstated, as they provide centralized resources, facilitate communication and collaboration, and enable the efficient management of data and applications across an organization.
The concept of "serverless" computing has emerged as a significant development in cloud computing, and it is important to understand what this term actually means. When managed services in the cloud are described as serverless, it does not mean that servers are entirely absent from the equation. Rather, it means that the responsibility for deploying, configuring, and maintaining the server infrastructure, whether physical or virtual, has been assumed by the cloud service provider. Organizations that utilize serverless offerings are relieved of the operational burden associated with server management, allowing them to focus on their core business objectives while the cloud provider handles the underlying infrastructure. This represents a significant shift in how computing resources are consumed and managed.
Servers fulfill a wide variety of roles within an organization, and these roles are as diverse as the needs of the organizations that deploy them. Authentication servers are responsible for verifying identities, which involves confirming usernames, passwords, and potentially additional verification codes that might be sent to a user's mobile device as part of multi-factor authentication processes. Network time synchronization servers ensure that clocks across the network remain synchronized, which is crucial for many applications and security protocols. File servers provide shared access to folders and files, enabling collaboration and centralized data storage. Database servers manage and provide access to structured data, while web servers host HTTP-based services that deliver HTML pages to clients. Application servers often combine the functionality of database and web servers to support complex applications. It is technically possible to run multiple server roles concurrently on a single server, although this practice is generally not recommended. Best practices suggest isolating server roles wherever possible to enhance security, performance, and manageability, though there are situations where consolidation may be appropriate.
Server Operating Systems
Unix and Linux represent two of the most significant and enduring families of server operating systems. Unix has been in existence for many decades and continues to be used today in various distributions, much like Linux. The fundamental difference between the two lies in their licensing models: Linux is open source, meaning its source code is freely available for modification and distribution, while Unix is proprietary, meaning it is owned and controlled by specific vendors. Despite this difference, Unix and Linux share many similarities. Both can run on a wide variety of hardware platforms, and both can be installed directly on hardware or embedded in firmware for specialized devices such as Internet of Things (IoT) devices. These operating systems are manageable through command-line interfaces, and they are case-sensitive, meaning that commands and file names must be entered with the correct capitalization. Graphical user interface (GUI) environments are also available for both Unix and Linux. The X Windows windowing system has long provided a graphical environment for interacting with Unix and Linux systems, either locally or remotely through a feature known as X forwarding. In Unix and Linux environments, file extensions are less important than they are in Windows. For example, whether a text file has a .txt extension or not has no functional relevance; the extension serves primarily as a convenience for users, helping them identify the type of content contained within a file.
A daemon in Unix and Linux is a background-running service that is not tied to any specific user account. Daemons run continuously to provide various system functions. For example, when running Docker application containers, the Docker daemon must be running at all times to manage and execute those containers on Unix or Linux systems. The concept of a daemon is fundamental to how Unix and Linux systems operate, providing essential services without requiring user intervention.
In the Windows environment, Microsoft Windows Server serves as the primary server operating system. Like Unix and Linux, Windows Server can be installed on physical hardware or within a virtual machine. It can also be embedded in firmware on certain devices. Windows Server can be managed through a command-line interface, which is not case-sensitive, unlike Unix and Linux. The command-line options include the standard Windows Command Prompt (cmd.exe) and PowerShell, a more powerful and flexible scripting environment. GUI management is also available for Windows Server, and both command-line and GUI options can be used locally or remotely. File extensions play a significant role in Windows environments. For instance, a .txt file is associated with a program such as Notepad, and double-clicking the file will open it in that application. In Windows, a service is a running program that is not tied to a user account and runs continuously whenever the machine is running. This is analogous to the concept of a daemon in Unix and Linux. Examples of Windows services include web server stacks and the SSH daemon that allows remote command-line administration on Linux systems.
On the macOS platform, the macOS Server application is available through the Mac App Store and requires macOS 11.3 or later to install. The macOS Server app provides centralized management of Mac and iOS devices, allowing administrators to apply device updates from a central location and manage devices through a web-based administration tool. This centralized approach to device management offers significant benefits in terms of efficiency and consistency.
Centralization Benefits
One of the primary advantages of using servers is the ability to centralize various aspects of information technology management. Centralized user accounts, file access, and database access mean that logging and auditing are also centralized, which greatly facilitates ensuring legal and regulatory compliance from a single location. While it remains necessary to monitor activity on client devices, centralization makes it considerably easier to track who did what and when. Centralized security permissions can be configured, for example, on a file server, rather than relying on workgroup computers that share folders and manage permissions sporadically across the network. In enterprise environments, centralized management is simply more sensible and efficient. Centralized backup is another natural benefit of server usage. Backing up user, group, and computer account configurations that are already stored on the server is facilitated because the data is already in a central location. Policies such as group policy and mobile device policy settings can be backed up, as can user data files and databases. Synchronizing data to the public cloud provides an additional layer of protection, allowing an on-premises file server to be replicated to the cloud as an offsite backup. In the event of a failure of the on-premises file server, a copy of the data remains available in the cloud.
BIOS Configuration
The Basic Input and Output System (BIOS) is a firmware instruction set that resides on a motherboard and also on expansion cards. The BIOS is responsible for the fundamental instructions that enable a server to start, check hardware, and boot an operating system from a specific point on a disk. When a physical server is powered on, the BIOS initiates the process of checking hardware and locating the operating system. The Complementary Metal-Oxide Semiconductor (CMOS) is an implementation of the BIOS in terms of its configuration. The BIOS refers to the general capabilities at the firmware level, while the CMOS represents the configuration of those capabilities, such as changing the boot order on a server. Because the BIOS is firmware, it is not immune to receiving updates. Firmware updates are available for motherboard BIOS chipsets and for BIOS chipsets on add-on cards, including RAID controllers. It is important to stay current with firmware updates to ensure the best stability and security.
One important consideration when configuring a BIOS is verifying that options for managing console redirection are available, which provides remote management capabilities. To enter the BIOS on a server, the typical procedure is to power on the server and press a specific keystroke, which varies depending on the BIOS manufacturer. If remote out-of-band control is available, meaning the operating system does not need to be running, BIOS configuration can be performed remotely without being physically present in a server room or at an equipment rack.
BIOS security is an important consideration. Setting a power-on password prevents the server from starting without someone entering the password. This is beneficial for security, but it also presents a trade-off because automatic server restarts would require someone to be present to enter the password, either physically or remotely through out-of-band management tools. A configuration change password can also be set, which does not need to be entered every time the machine powers on but must be entered before any BIOS settings can be changed. Every server should be configured with such a password to prevent unauthorized modifications.
The Trusted Platform Module (TPM) is a firmware solution that is normally built into server motherboards and can also be found in desktop and some laptop motherboards. It is available via an add-on expansion card or can be enabled as a virtual TPM with virtualization solutions such as VMware Workstation or Microsoft Hyper-V. The TPM provides cryptographic functionality, including ensuring boot integrity by verifying that operating system startup files are valid and have not been tampered with. It can also encrypt data at rest. Software solutions such as Microsoft BitLocker can work alongside the TPM, using a key stored in the TPM to encrypt disk volumes. This ensures that data at rest is protected locally on that server, and even if someone physically stole the server storage device, they would not be able to decrypt its contents.
The Preboot Execution Environment (PXE) is another option available at the hardware level in servers, desktops, and laptops. PXE allows a machine to boot to the network prior to booting the local operating system. This depends on setting the correct boot order in the CMOS, or it may require pressing a key such as F12 to initiate a network boot instead of a local operating system boot. PXE enables tasks that do not require a locally running operating system, such as pulling down a server operating system image, performing data recovery or forensic tasks, or conducting virus scanning without the host operating system running.
Several server BIOS settings are commonly encountered. A hardware RAID controller is often built into the server motherboard, and the appropriate keystroke can be pressed during the boot sequence to enter the hardware configuration for RAID and enable features such as disk striping, disk mirroring, or striping with distributed parity. The necessary storage media must be available; for example, disk mirroring requires at least two physical disks. Out-of-band remote server management is another common BIOS setting. Depending on the server, this may be built into the motherboard or require an add-on expansion card. Out-of-band management means the operating system does not need to be running to remotely control the server, unlike SSH for Unix and Linux hosts or Remote Desktop Protocol (RDP) for Windows servers. A special IP configuration, including a default gateway, must be configured to enable this capability. Asset and inventory tag information can also be enabled within the server BIOS, and inventorying tools can read this information and store it in a configuration database.
Configuring BIOS
A crucial part of server management is understanding the BIOS and its successor, the Unified Extensible Firmware Interface (UEFI). Both refer to the firmware-level instructions that enable a server to boot and control the behavior of its hardware components. Physical presence at the server is not necessarily required to access the BIOS hardware setup if remote management over the network has been enabled. For example, on a Dell PowerEdge server, the Integrated Dell Remote Access Controller (iDRAC) network cable can be plugged in and configured, allowing connection to the built-in web server for remote management even if the operating system is not installed or running.
In a sample BIOS simulator, navigation through the menu system at the top allows configuration of various settings. Under the Main menu, information about the system BIOS, including the date and version, is displayed. It is important to take note of these details to ensure the latest version is in use. Under the Advanced menu, submenus such as Processor Configuration, Integrated IO Configuration, Mass Storage Controller Configuration, PCI Configuration, Serial Port Configuration, and USB Configuration provide access to detailed settings. The USB Configuration allows determination of whether USB devices can be bootable and whether Legacy USB Support is enabled for older USB devices. Under the Security menu, passwords of various types can be set, including a Power On Password and a Front Panel Lockout setting that disables the power button on the front of the server when it is installed in a rack.
Under the Server Management menu, options such as Console Redirection and Baseboard Management Controller (BMC) settings are available. The BMC is firmware normally built into the motherboard for server-class hardware or enabled with an add-on card. It allows remote hardware-level access even if the operating system is not running and has its own Ethernet connector, typically an RJ45 connector, providing a separate network interface. The IP settings are configured accordingly, and the server can then be managed remotely, usually through an HTTPS interface built into the firmware. A built-in web server application in the iDRAC allows connection from a client-side web browser. A message about not trusting the HTTPS certificate is expected and can be bypassed. The Dell website provides a screenshot of the iDRAC interface, which allows management and monitoring of the server without an operating system running on the host. For security reasons, the network connection for remote console access should be on an isolated network or accessible only through a VPN.
The Boot Maintenance Manager provides options for changing boot settings, such as enabling network PXE boot before booting from the local hard disk. Settings can be saved and exited through the menu system. It is important to understand the server hardware and the type of BIOS it contains, keep the BIOS up to date not only for the motherboard but also for all add-on cards such as RAID controllers and baseboard management controllers, and document the configuration using enterprise-class tools that can inventory the configuration and save it externally.
UEFI and BIOS
The Unified Extensible Firmware Interface (UEFI) is a firmware standard that supersedes the traditional BIOS. While the BIOS is older and UEFI is newer, UEFI offers the same kinds of options as the BIOS along with additional features. At the server level, some servers, such as the HPE ProLiant Blade Server, provide multiple boot mode configurations, including Legacy BIOS mode and UEFI mode. Some options, such as secure boot, may require UEFI mode to be enabled.
The Preboot Execution Environment (PXE) allows a machine to boot to the network prior to the operating system booting. This is supported by both UEFI and the older BIOS, and the network interface card must support PXE boot. At the server level, network interface cards are usually built into the motherboard, often more than one. A remote management NIC may be built in, along with one or more standard NICs for standard network communication, and there may be slots for add-on cards. The PXE boot setting is configured in the CMOS boot order, and booting PXE before the local hard disk may be selected. However, physical presence at the server console or remote management at the hardware level is required to press the appropriate keystroke to execute the PXE boot. Network boot prior to operating system initialization may be used for malware scanning without the operating system running, applying an operating system image, or general troubleshooting.
UEFI offers several features beyond the BIOS. It supports mouse-based interaction with UEFI configuration options, not strictly keyboard-based input. It can boot from a GUID Partition Table (GPT) disk, which is a newer standard that supersedes the Master Boot Record (MBR). MBR supports four disk partitions on one physical disk, with each partition having a maximum size of approximately 2 terabytes. GPT removes these limitations, allowing very large disk partitions and, at the Windows level, at least 128 disk partitions on a single physical disk. IPv6 PXE boot is a UEFI feature that is supported when enabled in the motherboard UEFI configuration. UEFI also supports plug-in modules for modularity and additional functionality over time.
Secure boot is based on UEFI firmware and is designed to ensure that untrusted code does not execute before the operating system is up and running, where security checks are built into the operating system. Each file that comprises the operating system must be digitally signed by the software vendor for secure boot to work. The UEFI BIOS contains a collection of trusted digital certificates containing public keys that are used to validate the digital signature on each operating system file. If a file has been infected or tampered with, the signature will not be present, and when checked during secure boot against the certificates in the UEFI BIOS, execution will halt. Not every operating system supports secure boot; some Linux distributions do not. In some cases, secure boot must be disabled in the UEFI to allow installation of an operating system that does not support it, and a legacy mode boot operating system may need to be enabled in the UEFI settings. A screenshot of a UEFI configuration shows whether the secure boot option is enabled or disabled. In the Windows environment, the systeminfo command can be issued from the command line to obtain details about the operating system, including the BIOS version. Keeping the BIOS or UEFI up to date is important because firmware updates may address security problems, add new features, or correct stability issues.
Server Form Factors
Server hardware encompasses not only the components inside the server but also the size of the server case and the components that can fit within that case. Physical servers and their parts come in a variety of shapes and sizes, known as form factors. Not every hardware component will fit into every server. For example, add-on expansion cards that fit well in a server tower will not fit into a blade server, and a tower server cannot be rack-mounted in a server room rack. Form factors apply to computer casings, power supplies, motherboards, expansion cards, and other components. Ordering servers and physical hardware components requires careful attention to detail.
A tower server is a standalone computer that is not designed to be inserted into a rack. This is an older server form factor that does not scale well in a data center environment. Data centers are designed to house large amounts of computing equipment, including servers, storage arrays, and uninterruptible power supply (UPS) battery backup systems, typically organized into racks where equipment is installed vertically to save space. Tower servers would require shelving or similar arrangements to be stacked vertically, which is not their intended purpose. Rack-mountable servers are better suited for data center environments.
A rack-mount server is a thin server designed to be slid into tracks on a rack and secured with screws. The server slides in and out, and cables must be long enough to accommodate this movement when servicing or modifying the server at the hardware level. Most equipment racks in server rooms or data centers have front and back locking doors. The front door may be left unlocked while the back door, where cables are plugged in, is locked. Server equipment can be configured so that front-facing buttons are disabled, preventing someone from walking into the server room and pressing the power button to power off a server. Rack-mount servers greatly increase server density in a server room or data center because they are designed to slide into tracks within an equipment rack and stack one above the other. This allows a large number of rack-mount servers to fit in an equipment rack compared to the equivalent compute power of tower servers. Racks are also designed for cable management, with cable management arms or special conduits accommodating data, power, and network cables to allow flexibility. Rack-mounted servers are easy to install with special cases and rails, and they often come with rails and screws. Blinking LED lights on the front or back of rack-mounted server equipment can be used to identify a specific physical server, which is important when troubleshooting or dealing with a specific server among multiple racks containing hundreds of servers. The blinking light can also indicate that the server is being managed remotely.
A blade server is essentially a thin server circuit board that slides into a rack enclosure, similar to putting multiple server motherboards within what appears to be a single server case. The enclosure houses multiple blades or server circuit motherboards. Blade servers are designed to take the least amount of space possible, increasing server density in a server room or equipment rack even further than rack-mount servers. In a server room or data center, one or more equipment racks are typically tall and vertical, and the floor may need to be reinforced to accommodate the weight of the equipment, especially UPS battery packs. Equipment racks have front and back doors that can be locked, although cheaper racks may have no doors at all. Rack-mounted equipment is designed for larger-scale use, keeping equipment clean and organized and increasing the number of servers that can be accommodated compared to tower servers.
Calculating Rack Space Requirements
Calculating rack space requirements is an important consideration for server technicians. Server room and data center metal racks hold equipment stacked vertically, and these racks can become heavy depending on the equipment installed, such as UPS battery systems. The floor in the server room or data center must be constructed to support the weight, and tall racks must be properly balanced to prevent tipping and personal injury. When working with racks, equipment is slid into the rack with rail kits, which usually come with rack-mount servers along with screws, although in some cases they must be acquired separately.
The standard rack measurement is called a U, which is a unit of measurement for racks and rack-mountable devices. The U measures the amount of vertical space between the holes in a rack, which are the holes in the frame along the front and back of racks. One U equals 1.75 inches (1 3/4 inches). Two Us would be 3.5 inches (3 1/2 inches). This measurement is important when ordering equipment and racks to ensure that all equipment can be accommodated. Most devices installed in a rack-mountable environment have a vertical height between 1 and 7 Us, depending on the type of device. A server room rack might be 42U in height, although smaller racks are available that can be mounted on a wall to accommodate a small number of devices, such as network switches in a communications closet.
An equipment rack can contain many different types of equipment. In a typical example, blade server enclosures might be at the top, network switches and communication equipment in the center, and rack-mountable servers, storage arrays, or UPS battery systems below. Specific racks may be designated to accommodate only certain types of equipment for organizational purposes. In addition to the rack frame, which may have front and back locking mesh doors, a power distribution unit (PDU) provides power for everything on the equipment rack. Redundant PDUs are common and should be plugged into separate electrical circuits to increase resiliency against power failures. Redundant server power supplies should be plugged into separate PDUs so that if one circuit fails, the server remains running. Servers may also be plugged into an uninterruptible power supply (UPS), which contains batteries and is plugged into a different PDU.
Server rack components also include rack filler panels, which cover open spaces in the equipment rack to optimize airflow. Filler panels help channel cool air into the front of the equipment rack, where intake fans bring cool air into the system, and exhaust fans at the back of the equipment expel warm air. Filler panels help control hot and cold aisle airflow within a server room or data center. A cable management arm is another component, usually a folding component made of plastic or metal, that extends when rack-mounted equipment is slid out from the rack and contains the cables for that equipment. Cables must be long enough to allow the equipment to slide out for servicing or replacement. When purchasing a server rack, it is important to ensure that it can accommodate the equipment that will be installed. A wall-mountable server rack measuring 27Us vertically is an example of a locking cabinet, which is recommended for security reasons to provide an additional level of defense against unauthorized access to physical equipment. Server racks may come with rail kits and one or more exhaust fans to remove warm air from the back of the rack and funnel it into a hot aisle.
Central Processing Unit
The Central Processing Unit (CPU) is the brain of the server and comes on a physical chip. Server motherboards sometimes have multiple physical CPU slots, and it is important to match the correct type of CPU with the slot on the motherboard. To handle more intense workloads, adding more compute horsepower, meaning more CPU power, is called scaling up, which is a form of vertical scaling. Adding additional servers to handle a busy workload is called scaling out, or horizontal scaling. Scaling down is the opposite, reducing the amount of horsepower. In a virtual environment or the cloud, scaling is easily accomplished, and it is important not to deploy too much power in a virtual server because the cost is based on resource allocation. If 64 gigabytes of RAM and 12 CPU cores are not needed, the amount should be reduced to match workload requirements.
CPU architecture comes in two forms: 32-bit, which has a maximum addressable amount of memory (RAM) of 4 gigabytes, and 64-bit, which can theoretically address up to 16 exabytes of RAM. While 32-bit servers are older technology, they may still be encountered in some server rooms where older equipment remains in use. It is important to be aware of the limitations of 32-bit equipment, including the types of cards that can be plugged into slots and the maximum addressable amount of RAM. Modern servers are 64-bit based, meaning not only the CPU but also the motherboard and all buses interconnecting hardware components.
Virtual CPUs (vCPUs) can be configured for a virtual machine. In a virtualization environment, the number of virtual processors and the number of cores per processor can be specified. It is important not to always set the maximum number of processors and cores per processor, especially when running multiple busy virtual machines on the same physical hypervisor host. Virtual machines configured with excessive virtual CPUs and cores can slow down if they are waiting for an underlying physical CPU or core to become available. This is an example of vertical scaling for an on-premises virtual machine. In the cloud, virtual machine sizes (as they are called in Microsoft Azure) or instance types (as they are called in Amazon Web Services) determine how much is paid. More CPU compute power and more RAM mean more cost for every second the virtual machine is running. Different VM sizes have differing numbers of vCPUs, amounts of RAM in gigabytes, numbers of supported data disks, and maximum IOPS for disk throughput. All of these are related to scaling up and down, with the number of virtual CPUs being a focal point.
CPUs on mobile devices use the Advanced Reduced Instruction Set Computing (RISC) Machine, or ARM processor. ARM processors are used in devices such as Apple iPods, iPads, iPhones, Raspberry Pis, Microsoft Surface tablets, Android-based devices, and IoT devices. These processors are designed to provide appropriate compute power while drawing the least amount of power possible, especially for battery-powered devices. Hardware virtualization can be enabled at the hardware level, and the CPU must support it if a hypervisor and multiple virtual machine guests are to be run. On the Intel side, this is called Intel Virtualization Technology (Intel VT), and on the AMD side, it is called AMD Virtualization (AMD-V). CPU speed is measured in gigahertz (GHz), and the fastest possible CPUs that fit within budget should be selected. CPU cache memory is a small amount of very high-speed, expensive memory that speeds up the execution of CPU instructions expected to be called upon, increasing the CPU cache hit rate. Cache is used for recently used items, including CPU instructions, data, and anticipated instructions. References to L1, L2, and L3 cache may be seen when reading CPU details. L1 is the fastest, while L4, which is also available, is the slowest but usually comes in the largest amount because it is slower and less expensive. L1 cache should be a focus when selecting a CPU. CPUs must match the motherboard sockets. For example, an LGA 3647 socket, also called Socket P, would use an Intel Xeon W chip. A BIOS Virtualization Configuration screenshot shows the Advanced section where Intel Virtualization Technology can be enabled. If the server is going to act as a hypervisor, this option should be enabled.
Random Access Memory
A common myth about server performance is that adding RAM always improves performance. While this may be true to a certain point, it is not always the case. Random Access Memory (RAM) is electronic volatile memory used to run operating systems and programs and to work with data. It requires electricity to function and work with data, unlike disk storage devices, which are nonvolatile and retain data when power is cut off. Servers can accommodate different amounts and types of RAM, and motherboards have specific memory module placement rules regarding where chips are placed and how they must be inserted. It is important to refer to the specific documentation for the server motherboard model.
In an on-premises virtualization environment, a virtual machine's RAM can be configured. A slider allows the amount of RAM to be increased or decreased. Colored triangles to the right of the slider indicate the maximum recommended memory (blue), the recommended amount of memory (green), and the guest operating system recommended minimum (yellow). The guest operating system recommended minimum is determined based on the type of virtual machine and the operating system that will be installed. Some virtual machine environments allow dynamic memory to be enabled, which adjusts the amount of memory used while the VM is running. If a VM is set to use 16 gigabytes of RAM but currently only needs 4, the 12-gigabyte difference can be allocated to other virtual machines that need it at that time.
For a cloud-based server, the amount of RAM allocated is determined by the VM size. In Microsoft Azure, different VM sizes have differing amounts of RAM, along with the number of vCPUs, the number of supported data disks, and the IOPS value for disk throughput. More IOPS is better for disk-intensive workloads. For physical server RAM installation, a motherboard may require Error Correcting Code (ECC) memory. ECC memory chips can detect and fix memory errors at the bit level, using an extra parity bit. Most consumer-grade computing equipment uses non-ECC memory modules. Memory timing is another consideration. A motherboard might support RAM chip modules that support double pumping, which transmits data through the system from memory on both the rise and fall of a timing signal, such as that of the CPU measured in GHz. This increases throughput getting data into and out of the memory chips. Physical servers and their RAM chips require the right RAM chips to accommodate the slots on the motherboard, such as Double Data Rate (DDR) RAM. There are many versions of DDR, including DDR3, DDR4, and DDR5. A server motherboard designed for DDR3 chips might only accept DDR3 memory chips, so it is important to refer to the motherboard documentation. The maximum amount of supported RAM and the number of pins in the memory modules, such as 240 pins, must also be identified. Server memory is normally ECC-based, and an example of ordering server ECC RAM is 32 GB DDR4 modules.
Buses and Interfaces
Motherboards and expansion cards in a server need a way to move data around, which is where the bus comes in. Computer buses are analogous to a real bus that moves people around using roads, with the difference being that a computer bus moves data bits around the system using a variety of different buses. Hardware components interface with these buses through various types of cards and connectors.
Peripheral Component Interconnect (PCI) is an older, legacy expansion card slot standard that became a standard in the 1990s. PCI supports 32/64-bit and up to 800 Mbps of throughput. PCI slots on a motherboard allow PCI expansion cards to be inserted, such as network interface cards, graphics cards, or other add-on cards. Modern motherboards are unlikely to have PCI slots, but server technicians may still encounter older equipment that needs to be supported. PCI-X stands for PCI-extended and is a slot and card standard that supersedes the original PCI standard. PCI-X supports 64-bit parallel transmission and has up to 4 Gbps of throughput. It is important to know whether a server motherboard can accommodate PCI-X cards when replacing a failed add-on card.
PCI Express (PCIe) is the standard most likely to be encountered with modern server motherboards. It supersedes PCI and PCI-X and allows for more simultaneous communication channels, meaning more speed through the PCIe bus to and from cards. For example, two communication channels (x2) versus 16 communication channels (x16) provide much more throughput. An x2 card would equate to 500 Mbps, while a 16-lane PCIe card and slot would allow up to 15 Gbps of throughput. An x8 card can be plugged into an x16 slot and will function properly, so an x16 slot does not require an x16 card.
Server expansion slots are important, and the number of expansion slots available on a server motherboard must be identified. When ordering new server motherboards, it is important to consider how expansion cards will be accommodated and to identify the types of slots. Some server motherboards have different types of slots on the same motherboard, mixing PCIe and PCI-X. USB connectors may be available not only on the external back or front of a server but also through a PCIe expansion card that provides additional USB slots. The USB 1.0 standard was 12 Mbps, USB 2.0 was 480 Mbps, USB 3.0 was 5 Gbps, and USB 4.0 supports up to 40 Gbps. Expansion cards that might be inserted in a server slot include RAID storage controllers (if not built into the server motherboard), network interface cards (NICs), Host Bus Adapters (HBAs) for connectivity to a fibre channel storage area network (SAN), and remote server management or baseboard management controller (BMC) cards that allow remote server management at the hardware level even if the operating system is running.
Servers and Electricity
Servers need power, even virtual servers, because they run on a physical hypervisor host that needs to be powered. There are several considerations when it comes to servers and electricity. Power connectors include the standard NEMA connector, which is the standard power plug and receptacle used in North America and other countries adhering to the North American standard. NEMA connectors normally have three prongs, one of which is used as a ground. An Edison connector is similar but more rounded, also with three prongs. Twist-lock connectors are used for industrial-grade equipment to prevent accidental unplugging from the socket.
Electricity has attributes such as voltage (V). When there are two points in a circuit with different potential electrical charges, the flow of electrons between them causes electricity to flow. A greater difference between the two electrical charges results in a faster or greater flow of electrons. This can be dangerous when the human body becomes part of a circuit, as contact with sensitive electrical components at a different charge can result in a rush of electrons that can damage components. Static charge can build up from low humidity, walking over carpet, and other factors, and can damage electrical components without any visual indication. Electrostatic discharge (ESD) will be covered in more detail later.
Voltage can be alternating current (AC), which has a tiny fluctuation in the current supplied from the power grid, or direct current (DC), which has the same flow of electricity at the same constant voltage, normally provided by batteries. In North America, the standard voltage is between 110 and 120 volts, while in some other parts of the world, such as Jamaica and parts of Europe, 220 to 240 volts is the norm. Some equipment can reduce voltage for equipment that requires a smaller amount. Older power supply units (PSUs) may have a switch to change between 110 and 220 volts, while modern equipment should detect the difference in AC flow and adjust accordingly.
The relationship between power draw and CPUs in a server is important. The PSU in a server normally provides +5V and +12V connectors. A Voltage Regulator Module (VRM) built into the firmware ensures components only get the voltage they need. CPUs do not normally need as much power as is fed directly from the PSU, so the VRM ensures the CPU gets only the voltage it actually needs. Industrial environments requiring robotic machines, motors, and hydraulics may need different power requirements, such as 440 or 480 volts. Most household and business environments use one-phase power, while industrial environments normally use three-phase power for higher voltages.
The PSU in a server must provide enough power to accommodate all devices plugged into the server, including network interface cards, disks (each disk draws power to drive motors if it has moving parts, such as hard disk drives, while solid-state drives have less power draw because they have no moving parts), firmware, CPUs, and RAM chips. The power draw of these components is documented in the literature, and the total power draw should not exceed the PSU listed wattage. PSU wattage varies depending on the equipment, such as 800W or 1600W. Most servers have dual PSUs for redundancy, and the PSU must fit within the server enclosure, with its own form factor determining what it will fit into.
Uninterruptible Power Supplies (UPSs) are rack-mounted and are collections of batteries that are continually charged while connected to the power grid. Equipment is plugged into the UPS, which has many plug sockets. When the power grid fails, the UPS provides power to plugged-in equipment, allowing at least a graceful shutdown. UPS power draw must be calculated carefully. The normal component power draw is determined by dividing the wattage by the server power factor ratio, normally 0.9, to get a Volt-amp (VA). The VAs for all devices in the machine are added up and multiplied by 1.3 to account for power draw fluctuations. The UPS VA listing must be more than the sum of the VA draw for all server components and all servers plugged into the UPS.
Managing Electrostatic Discharge
In a perfect world, the human body and all clothing and jewelry would have the exact same charge as the electrical components being handled, preventing the flow of electrons and preventing damage to equipment. Electrostatic discharge (ESD) occurs when there is a difference in electrical potential between two endpoints, causing a rush of electrons. Sensitive electrical components such as motherboards, circuit boards, add-on cards, RAM chips, and CPU chips are easily damaged by ESD through simple touching, and there is often no visual indicator that the equipment has been damaged.
Preventative measures include storing spare components in anti-static bags. When handling hardware components, the question of whether the Power Supply Unit should be plugged in during servicing arises. Leaving the machine plugged in provides a proper earth ground to a grounded power outlet, allowing built-up static charges to dissipate. Most PSUs have an on/off switch that can be flipped off while still being plugged into a grounded outlet. However, unplugging the computer and ensuring proper ESD mechanisms are in place is ideal for utmost safety. Proper ESD mechanisms include an ESD bracelet or strap, which goes around the wrist and connects to an alligator clip that can be clamped onto unpainted metal, such as the server chassis, to put the body at equal electrical charge with what is being touched. ESD footwear and ESD mats are also available. ESD mats are rubber mats that are grounded and stand on while servicing sensitive components. Entire workbenches designed to dissipate and prevent ESD are available in shops where sensitive components are handled frequently. Keeping air humidity at approximately 50% also helps prevent ESD, as low humidity increases the possibility of ESD.
Server Environmental Considerations
Physical servers are affected by the environment in which they exist. Temperature is a critical factor. Server rooms and data centers should not be too warm, as this is bad for electrical equipment. When a CPU gets too warm, it may throttle itself, slowing down to draw less power and generate less heat, but this degrades performance. Equipment may fail or shut down if it becomes too warm. Humidity levels should be around 50%, as too dry air can increase the likelihood of electrostatic discharge. Heating, ventilation, and air conditioning (HVAC) is a major consideration for server rooms and data centers.
Hot aisles and cold aisles are about feeding cool air to equipment. Cool air is fed through the equipment by intake fans, keeping the temperature cool inside the server. The exhausted air, which has become warm, goes into a hot aisle and is removed. Cool air is brought in continuously, and hot exhaust air is removed, either filtered outside the facility or into an air conditioning unit to be cooled again. Larger server rooms and data centers have many racks of heat-generating equipment, making it very important to keep hot air separate from cool air. Air containment panels, which can be metal or rubber curtains, are installed to keep hot and cool air separate and control airflow.
Properly designed server rooms and data centers often have raised floors with perforated floor tiles. The air conditioning unit feeds cool air under the floor and up into the front of equipment racks where intake fans bring it into the equipment. Warm exhaust air comes out of the back of the equipment racks into a hot aisle and is fed up and around to an air conditioning unit or expelled outside. Server rack filler panels are used to cover vacant spots in a server rack to optimize airflow. Fire suppression is another environmental consideration. Periodic fire drills should be conducted so that staff are aware of what to do in a fire emergency. Server rooms, office spaces, and data centers must be equipped with smoke detectors that are tested periodically. Class C fire extinguishers should be used for electrical fires. Water should never be used on an electrical fire because water conducts electricity, so water sprinkler systems should not be used in server rooms or data centers for putting out electrical equipment fires.
Conclusion
Servers are fundamental to modern information technology infrastructure, providing centralized resources, enabling efficient management of data and applications, and supporting a wide range of critical functions. Understanding the various aspects of server usage, from operating systems and BIOS configuration to form factors, power requirements, and environmental considerations, is essential for anyone involved in server administration or support. By applying the knowledge and best practices outlined in this document, server technicians and administrators can ensure that their server environments are stable, secure, efficient, and well-managed.