Unit 1 - Practice Quiz

Before any installation, it's essential to plan what the server will be used for (e.g., file server, web server) and determine the necessary CPU, RAM, and storage to support that role.

RAID is a technology that combines multiple physical disk drives into one or more logical units for the purposes of data redundancy, performance improvement, or both.

The BIOS (Basic Input/Output System) or its modern successor, UEFI (Unified Extensible Firmware Interface), is the firmware used to perform hardware initialization during the booting process.

Server Core is a minimal installation option for Windows Server that includes most but not all server roles. It has a smaller disk footprint and attack surface because it lacks the standard graphical user interface (GUI).

Ethernet cables, such as Category 6 (Cat 6), are the standard for wired networking in local area networks (LANs) to connect devices like servers, computers, and switches.

A Network Interface Card (NIC) is a hardware component that connects a computer to a computer network.

DHCP is a network management protocol used to automate the process of configuring devices on IP networks, thus allowing them to use network services such as DNS, NTP, and any communication protocol based on IP.

Hyper-V is Microsoft's native hypervisor technology for creating and managing virtual machines (VMs) on Windows Server.

A virtual switch acts like a physical network switch, allowing virtual machines to communicate with each other, with the host machine, and with devices on the external physical network.

Proper documentation helps other administrators understand the setup, makes troubleshooting easier, and is critical for rebuilding the server in a disaster recovery scenario.

The HCL is a list of computer hardware (typically, components) that is compatible with a particular operating system. Using compatible hardware helps ensure a stable and reliable installation.

Microsoft requires a valid product key or another form of licensing (like volume licensing) to activate the operating system, which verifies that the copy is genuine.

DNS acts like the phonebook of the internet, translating domain names like 'www.google.com' into the numerical IP addresses that computers use to communicate with each other.

A virtual machine is a software program that emulates a complete computer system, allowing an operating system to be installed and run on it just as if it were a physical machine.

The physical setup, including mounting the server in a rack and connecting power, is the foundational step before any software configuration can begin.

A patch panel is essentially a static switchboard that helps terminate and organize cabling, making it easier to manage, troubleshoot, and make changes to the network layout.

NIC Teaming, also known as load balancing and failover (LBFO), allows multiple network adapters to be grouped together to increase throughput and provide network redundancy.

IPv4 addresses are 32-bit numbers represented as four decimal numbers separated by periods. IPv6 addresses are much longer and use a hexadecimal format.

An External virtual switch binds to a physical network adapter on the host, allowing virtual machines to communicate with devices on the external physical network.

The installer requires a hard drive to be partitioned (divided into sections) and then formatted with a file system (like NTFS) to create a volume where the operating system files can be stored.

While all options are part of planning, the core function of a file server is data storage and access. Therefore, defining the storage redundancy (RAID level) and a comprehensive backup/recovery strategy is the most critical initial step to ensure the primary goals of data protection and high availability are met.

For deploying multiple identical servers, using a customized image is far more efficient than manual methods. It automates the installation of the OS, drivers, roles, and features, ensuring each server is configured identically and drastically reducing the time and potential for human error associated with manual setups.

RAID 10 combines the performance of RAID 0 (striping) with the redundancy of RAID 1 (mirroring). It offers excellent read and write performance without the write penalty found in RAID 5, making it ideal for I/O-intensive applications like databases. RAID 5 suffers from slow write performance, and RAID 0 offers no redundancy.

Data center racks are designed for front-to-back airflow (cold aisle/hot aisle). Empty slots disrupt this design, allowing hot exhaust air to circulate back to the front intake. Blanking panels block these gaps, forcing cool air through the servers and maintaining the proper airflow pattern, which is crucial for efficient cooling.

The installation option without "(Desktop Experience)" is known as Server Core. It installs a minimal version of the OS without a graphical user interface (GUI). This results in a smaller disk footprint, lower resource usage, and a reduced attack surface, making it the ideal choice for remotely managed infrastructure servers.

The Windows Server installation media includes many common drivers, but not all of them, especially for newer or specialized hardware RAID controllers. When the OS setup cannot see the logical disk presented by the controller, it's necessary to manually provide the driver for that controller using the 'Load driver' feature during the installation process.

Category 6 (Cat 6) cabling can support 10 Gbps, but only up to 55 meters. Category 6a (Augmented) is designed to support 10 Gbps Ethernet (10GBASE-T) reliably up to the full 100-meter distance. Cat 5e is limited to 1 Gbps, and while Cat 7 has stricter specifications, Cat 6a is the recognized standard for this requirement.

A crossover fault indicates that the pinout on one end of the cable is wired for a crossover connection (e.g., pins 1 and 2 are swapped with 3 and 6) while the other end might be wired as straight-through, or both are wired incorrectly. It specifically refers to the swapping of the transmit (TX) and receive (RX) pairs, creating an unintended crossover cable.

In most NIC Teaming load balancing modes, including Address Hash, a single communication stream (or session) between two endpoints will only use one of the NICs in the team. The teaming provides an aggregate bandwidth of 2 Gbps for multiple, concurrent sessions to different endpoints, but any single session is limited to the speed of a single link, which is 1 Gbps.

LACP is the IEEE standard (802.1ax, formerly 802.3ad) for dynamically creating link aggregations between a server and a switch. It requires configuration on both the server and the switch. This mode provides superior load balancing and failover detection compared to static or switch-independent modes and is the best practice when supported by the switch.

The subnet mask 255.255.248.0 (or /21) defines a network where the IP addresses can range from 172.20.96.0 (Network ID) to 172.20.103.255 (Broadcast Address). The server's IP address 172.20.100.5 is within this range. The default gateway must be on the same subnet. Of the options, only 172.20.100.1 falls within the valid host range of 172.20.96.1 to 172.20.103.254.

The ability to resolve domain names to IP addresses is handled by the Domain Name System (DNS). If clients cannot browse by name but can by IP, it's almost always a DNS resolution problem. The DHCP server must be configured with Option 006 DNS Servers to provide the IP addresses of the DNS servers to the clients.

Generation 2 virtual machines are designed for modern 64-bit operating systems. They use UEFI firmware instead of the legacy BIOS used by Generation 1. This UEFI support is what enables features like Secure Boot, PXE boot using a standard network adapter, and booting from a SCSI virtual hard disk.

In Microsoft Hyper-V, a point-in-time image of a virtual machine's state, data, and hardware configuration is called a 'Checkpoint'. Checkpoints allow you to revert the VM to a previous state, which is very useful for testing updates or configuration changes. 'Save State' simply hibernates the VM, and 'Backup' implies a more robust, long-term data protection mechanism.

An 'Internal' virtual switch creates a network path between virtual machines on the host and the host itself. It does not, however, connect to any physical network adapter, thus providing the required isolation from the external network. A 'Private' switch only allows VM-to-VM communication (not with the host), and an 'External' switch connects to the physical network.

Checking this box is the standard configuration. It unbinds the TCP/IP stack from the physical NIC and binds the virtual switch protocol instead. Then, it creates a new virtual network adapter in the host (management) OS and connects it to the virtual switch. This allows the host to share the physical NIC and communicate on the network along with the virtual machines connected to that switch.

This configuration provides the highest level of redundancy. It protects against PSU failure, UPS failure, and the failure of a single electrical circuit or breaker. Connecting to the same UPS or circuit means that a failure of that single point would still take the server offline.

The operating system installer sees logical drives, not individual physical disks, when a hardware RAID controller is used. You must first boot into the controller's own BIOS/utility (e.g., by pressing Ctrl+R during POST) to define the RAID level and create the array (logical drive). Without this step, the OS installer will not have a target volume to install to.

The subnet mask 255.255.255.224 is a /27 network. This mask means the subnets increment in the last octet by 32 (256 - 224 = 32). The network IDs are therefore .0, .32, .64, .96, .128, .160, .192, .224. The IP address 10.10.5.200 falls into the range that starts at 10.10.5.192. Therefore, 10.10.5.192 is the network ID.

The VHDX format is the successor to VHD and is the default for new VMs in modern versions of Windows Server. It supports virtual disks up to 64TB (compared to 2TB for VHD), includes internal logging to improve resilience against corruption, and generally offers better performance.

This is a trick question, as you cannot have a PDC emulator in each site (it's a single-master role per domain). The best practice is to keep the forest-wide roles (Schema Master, Domain Naming Master) together at the primary/most stable site (New York). The domain-wide roles (PDC, RID, Infrastructure) should also be kept together on a single, well-connected DC to reduce inter-role replication latency. Placing them in London offers a good balance of latency for both New York and Tokyo. Keeping all roles together on one server is crucial for performance and consistency.

The write penalty is the number of I/O operations required for each single write request from the host. For RAID 10 (a mirror of stripes), a write requires writing the data to two disks (one in each side of the mirror), so the penalty is 2. For RAID 6 (dual parity), a small write operation requires reading the old data, reading the old P parity, reading the old Q parity, then writing the new data, writing the new P parity, and writing the new Q parity. This results in a total of 6 I/O operations, giving it a write penalty of 6. Therefore, RAID 10 is significantly better for write-intensive workloads.

For RDMA to function correctly through a SET team for guest VMs, the underlying physical NICs must have RDMA enabled prior to the team's creation. If the team is created first, the RDMA capability is not properly exposed to the virtual switch. The team must be destroyed, RDMA enabled on the physical adapters (e.g., using Enable-NetAdapterRdma), and then the SET team and vSwitch recreated. The load balancing algorithm does not directly affect RDMA functionality in this manner.

While the folder structure is used for adding files to the final Windows installation, Windows PE (the pre-installation environment) does not automatically search it for drivers. To load a critical boot driver like a storage controller, you must explicitly tell Windows PE where to find it. This is done by adding a <DriverPaths> entry within the Microsoft-Windows-PnpCustomizationsWinPE component during the windowsPE configuration pass in the autounattend.xml file.

Switch Independent teaming with Dynamic load balancing is designed for distribution of traffic, but it hashes outbound traffic based on a 2-tuple (source/destination IP) or 4-tuple hash (IPs and ports). This means a single, large file transfer session between two specific endpoints will always be hashed to the same team member NIC. Therefore, its maximum speed is limited to that single 10GbE link. It provides aggregate bandwidth for multiple clients but does not aggregate bandwidth for a single session. The perception of 1GbE-like speeds could be due to other bottlenecks, but the fundamental limitation described is the most likely networking-specific cause.

When DHCP servers in a failover relationship lose contact, they continue to serve their portion of the addresses for the duration of the MCLT. Once the MCLT expires, each server assumes the other is completely offline. They transition to a 'Partner Down' state and each server will take authoritative control of 100% of the address scope to ensure service continuity. In a network partition scenario where both servers are still active, this will inevitably lead to IP address conflicts as both servers may offer the same IP address to different clients.

Generation 2 VMs use UEFI firmware. When Secure Boot is disabled, the UEFI firmware attempts to find a bootloader at a standard, default fallback path, which is \EFI\Boot\bootx64.efi. If the custom Linux installation placed its bootloader (e.g., GRUB) at a different path (like \EFI\ubuntu\grubx64.efi) and did not register it in the UEFI boot variables, the firmware will not find it and fail to boot. The solution is to either rename/copy the bootloader to the default path or use the UEFI shell to manually launch it from its correct location.

Alien Crosstalk (ANEXT) is a critical performance parameter for 10GBASE-T, representing the noise induced on one cable from adjacent, parallel cables. While a single Cat 6a UTP cable can support 10GBASE-T up to 100 meters, its susceptibility to ANEXT increases significantly when tightly bundled with other cables carrying high-frequency signals. This interference can corrupt the signal and cause the link to become unstable or down-negotiate. Using shielded (F/UTP or S/FTP) cabling or maintaining separation between bundles is the standard mitigation for ANEXT.

CPU C-states are idle power-saving states. When a core is in a deep C-state (like C6), it takes a non-trivial amount of time (many microseconds) for it to 'wake up' and return to an active (C0) state to process an incoming packet or interrupt. This wake-up time introduces jitter and unpredictable latency. For ultra-low latency applications, it's standard practice to disable all C-states in the BIOS, forcing the CPU cores to always remain in the C0 state, ready to execute instructions immediately at the cost of higher power consumption.

Unlike full Server or Server Core, Nano Server has an 'offline' servicing model. All roles, features, and packages (like the DNS server role or antivirus) must be embedded into the VHDX image during its creation using the New-NanoServerImage cmdlet. You cannot use Server Manager or Install-WindowsFeature to add roles to a running Nano Server. This makes the initial image creation a critical planning step, as any forgotten package requires building and deploying an entirely new image.

The entire purpose of the DNS Socket Pool is to randomize the source UDP port used for DNS responses, making it much harder for an attacker to guess the parameters needed for a cache poisoning attack. A standard DNS query is sent to destination port 53, and the response comes from source port 53. With Socket Pool enabled, the response comes from a random high-numbered port in the pool. Older or poorly configured network devices or firewalls may have overly restrictive rules that expect all DNS traffic to originate from port 53 and will drop these legitimate, randomized-port responses, causing a timeout.

SR-IOV works by creating a direct path (a Virtual Function or VF) between the VM and the physical NIC, bypassing the Hyper-V virtual switch processing layer on the host. Because the data path is offloaded to the hardware, any policies or features that are implemented in the software of the virtual switch (like QoS, Port ACLs, traffic monitoring, etc.) are not applied to the SR-IOV traffic. There is a trade-off between the raw performance of SR-IOV and the rich feature set of the Hyper-V virtual switch.

Dynamic Memory works by starting the VM with the specified 'Startup RAM'. After the OS boots, the Hyper-V integration services (specifically the balloon driver) communicate memory pressure to the host, which then allocates more memory up to the maximum. However, many applications like Java or SQL Server try to reserve their full configured memory (the 4GB heap) at the moment they launch. If the currently allocated memory (which starts at 2048 MB) is less than what the application requests at that instant, the allocation fails immediately. The application doesn't wait for the dynamic memory mechanism to respond. The solution is to set the Startup RAM to a value greater than or equal to the application's fixed memory requirement.

The bcdboot command copies boot environment files from the Windows image and configures the Boot Configuration Data (BCD) store. The command bcdboot <source_windows_directory> needs to know where to write the boot files. By default, it writes to the active system partition. In a complex manual setup, you must explicitly tell it which partition to use. The correct syntax should specify the source and also the destination system partition, for example: bcdboot E:\Windows /s S:, where S: is the letter assigned to the physical disk's EFI or System partition. Without the /s switch, it may fail to find the correct location or write to the wrong one.

Node Interleaving stripes memory addresses across all memory controllers in the system. This creates a single, uniform memory access (UMA) domain. While this might benefit some niche high-performance computing applications that need massive memory bandwidth, it is detrimental to virtualization. A NUMA-aware hypervisor like Hyper-V tries to keep a VM's CPU (virtual processor) and its memory on the same physical NUMA node (socket) to ensure low-latency local memory access. Enabling Node Interleaving prevents this, forcing CPUs to frequently access 'remote' memory attached to other sockets, which significantly increases latency and reduces overall performance.

VMQ is designed to offload packet sorting for virtual machines directly to the NIC hardware, delivering packets to specific VM queues. RSS is for distributing the host's (or a single VM's) network processing across multiple CPU cores. On many NICs, the base CPU core for VMQ is Core 0. Similarly, the base core for RSS might also be Core 0. When both are active, the hardware may prioritize processing the VMQs, leaving the host's own network traffic to be processed largely by that same base CPU core. The fix is to use PowerShell cmdlets like Set-NetAdapterVmq and Set-NetAdapterRss to configure different base processors and a different range of processors for VMQ and RSS to ensure they do not overlap and conflict.

Multi-fiber MPO/MTP systems for technologies like 40GBASE-SR4 use parallel fibers for transmitting and receiving (e.g., 4 fibers to transmit, 4 to receive). For a connection to work, the transmit (Tx) fibers on one end must align with the receive (Rx) fibers on the other. A 'Type A' cable maps fiber 1 to pin 1, 2 to 2, etc. A 'Type B' cable reverses the pins (1 to 12, 2 to 11, etc.). A direct connection between two transceivers requires one crossover in the path. If you use a Type A cable, you are connecting Tx to Tx and Rx to Rx, so the link will not establish. You need a Type B cable to provide the necessary Tx/Rx flip.

The entire purpose of temporary IPv6 addresses is to provide a layer of privacy for outbound connections, preventing a host from being tracked over time by its stable IPv6 address. Registering this short-lived, randomized address in a public service like DNS would defeat that purpose. Therefore, the default behavior of the Windows networking stack is to perform dynamic DNS registration only for its stable public IPv6 addresses, not the temporary ones. This is by design and is the correct, expected behavior.

For networking to function in a nested virtualization environment, the Layer 2 traffic from the nested VMs must be able to pass through the parent VM's virtual NIC. The vSwitch on the physical host treats the parent VM's vNIC as a single endpoint with a single MAC address. However, the nested VMs will have their own, different MAC addresses. By default, the vSwitch will drop this traffic. You must enable MAC Address Spoofing on the parent VM's vNIC settings. This tells the physical vSwitch to allow multiple MAC addresses to originate from that single port, enabling the nested vSwitch to function correctly.

When using Shared LOM mode, both production traffic from the OS and management traffic from the iDRAC/iLO flow through the same physical NIC and cable. To segregate this traffic, the physical switch port must be configured as a trunk port (allowing multiple VLANs, e.g., 10 and 100). Crucially, the iDRAC/iLO firmware itself must then be configured to tag its own traffic with the management VLAN ID (100). This allows the switch to differentiate the management packets from the production packets (which would be tagged with VLAN 10 by the OS's NIC driver) and direct them to the correct network segment. Simply setting a static IP is not enough to achieve layer 2 isolation.