Unit 2 - Practice Quiz

Changing the default, often generic, server name to a meaningful one is a fundamental first step for easy identification and management on a network.

Setting the correct time zone, date, and time is crucial for log correlation, scheduled tasks, and authentication services like Kerberos, which can fail if time skew is too great.

A server role is a collection of software components that allows a server to perform a specific function, such as serving web pages (Web Server role) or managing user accounts (Active Directory role).

The Web Server (IIS) role provides the necessary services to host websites and web applications. IIS stands for Internet Information Services.

DHCP is used to automate the process of configuring devices on IP networks, allowing them to use network services without requiring manual configuration.

DNS acts like the phonebook of the internet, resolving domain names into the numerical IP addresses necessary for locating and identifying computer services and devices.

While updates can also fix bugs and add features, their most critical function is to patch security holes that could be exploited by malicious actors.

A patch is a software update specifically created to correct issues, improve performance, or, most commonly, fix a security vulnerability in an existing program or operating system.

SSH is the standard for secure remote administration, providing an encrypted connection to prevent eavesdropping. Telnet is an older, insecure alternative.

RDP is a proprietary protocol from Microsoft that allows a user to connect to another computer over a network and interact with its graphical desktop.

An SLA is a formal agreement between a service provider and a client that specifies metrics like uptime (e.g., 99.9% availability), response times, and penalties for failure to meet the targets.

CPU Utilization measures how busy the central processing unit is. Consistently high utilization can indicate that the server is overworked and may need an upgrade or load balancing.

When a server runs out of physical RAM, it starts using the much slower hard disk as 'virtual memory' (a process called swapping or paging), causing high disk activity and slowing down the system.

Capacity planning involves analyzing current usage trends to predict future requirements for CPU, memory, storage, and network bandwidth to avoid performance issues.

Primary storage is where the operating system, applications, and frequently accessed data reside. SSDs and HDDs are the most common forms of primary storage in servers.

While both serve files, a NAS is a purpose-built appliance with a streamlined operating system designed solely for providing file-level storage services on a network.

A SAN presents storage to servers as if it were locally attached disks (raw blocks). This is different from NAS, which presents storage as network file shares.

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

RAID 1, or mirroring, creates an exact duplicate of data on a second drive. If one drive fails, the system can continue operating using the mirrored copy, ensuring data redundancy.

RAID 0 splits data across multiple disks, allowing for faster read and write operations because multiple drives work in parallel. However, it offers no fault tolerance; if any disk fails, all data is lost.

The w32tm command is the primary tool for configuring the Windows Time service. This specific command sets the manual peer list to the specified NTP server and updates the configuration, forcing the server to synchronize with that source. The other options configure the time zone, computer name, and IP address, which are different local properties.

Host Header Bindings (or Host Names) allow a single web server with one IP address and port (like port 80) to serve multiple websites. IIS inspects the HTTP host header in the client's request to determine which specific site to serve the content from.

An Exclusion Range is the correct feature to prevent the DHCP server from handing out a specific, contiguous block of IP addresses from within a defined scope. This is used specifically for addresses that are already assigned statically. Reservations are for assigning a specific IP to a specific MAC address.

WSUS is designed for centralized update management. By creating different computer groups (e.g., 'Test Servers', 'Production Servers'), an administrator can approve updates for the test group first. After verifying stability, the same updates can then be approved for the production group, fulfilling the requirement for a staged and controlled rollout.

SSH is the industry standard for secure, encrypted command-line access. It is native to Linux and fully supported on modern Windows Servers, making it the perfect choice for managing heterogeneous environments from a single workstation. RDP and VNC are GUI-based, and Telnet is insecure as it transmits data in plain text.

To calculate the permissible downtime, first find the total minutes in a 30-day month: . The allowed downtime is . Therefore, the total permissible downtime is minutes.

A high Avg. Disk Queue Length (generally, a sustained value over 2 per spindle is a concern) indicates that I/O requests are waiting to be processed by the disk subsystem. Since CPU utilization is low, the processor is not the bottleneck. The disk cannot keep up with the demands from the application, pointing directly to a storage I/O bottleneck.

First, convert the free space to a consistent unit: 2 TB is equal to 2048 GB. To find the number of months until the storage is full, divide the total free space by the monthly growth rate: months. The closest answer is 8 months.

A SAN is designed specifically for this use case. It provides block-level storage access over a dedicated network (like Fibre Channel or iSCSI), which is ideal for clustered applications like databases that require high-speed, shared access to raw storage volumes. NAS provides file-level access, and DAS is not shared.

iSCSI is a storage networking protocol that works on top of TCP/IP. Its primary advantage is the ability to run on standard Ethernet hardware (switches, NICs, and cabling), making it a cost-effective alternative to Fibre Channel, which requires its own specialized and more expensive infrastructure.

RAID 6 meets the requirement perfectly. It uses double parity (or an equivalent mechanism) to provide fault tolerance for up to two simultaneous drive failures. RAID 5 can only handle a single drive failure. RAID 10 (with 6 disks) could only survive two failures if they occurred in different mirrored pairs. RAID 1 would not maximize capacity.

The DHCP renewal process begins at the T1 timer, which by default is set to 50% of the lease duration. At this point, the client attempts to contact the DHCP server that issued the lease to renew it. If that fails, it will try again at the T2 timer (87.5% of the lease duration).

JEA is a security technology designed for the principle of least privilege. It allows you to create constrained, role-based administration endpoints where non-administrators can run specific commands, scripts, and executables as a temporary, privileged virtual account, without giving them broad permissions.

Availability is calculated using the formula: . Plugging in the values: . This corresponds to 99.0% availability.

In a RAID 5 configuration, the capacity of one of the disks is used for parity information, providing single-disk fault tolerance. The formula for usable capacity is (N-1) * S, where N is the number of disks and S is the size of the smallest disk. In this case, it is .

A performance baseline captures the metrics of a system during its normal operational state. This baseline is crucial for capacity planning because it provides the foundation for trend analysis (forecasting when resources will be exhausted) and anomaly detection (identifying when current performance deviates from the norm).

This is a proactive alert because it warns the administrator of a potential issue (running out of disk space) well before it causes a system failure or service disruption. This allows them time to intervene and correct the problem. The other options are reactive, as they trigger after a failure has already occurred.

For any change to a critical production system, having a reliable and tested rollback plan is paramount. A pre-patch backup or snapshot allows the administrator to quickly revert the system to its previous state if the patch causes unexpected issues, thereby minimizing downtime and risk.

SSDs excel at random I/O operations because they have no moving parts and offer extremely low latency for accessing data, regardless of its physical location on the drive. This makes them ideal for OLTP database workloads, which are characterized by high rates of random reads and writes. HDDs suffer from seek time latency, which significantly slows down random I/O.

Windows Deployment Services (WDS) is the server role specifically designed for network-based deployment of Windows operating systems. It works with Pre-boot Execution Environment (PXE) to allow client machines to boot from a network adapter and then install an OS image stored on the WDS server.

RAID 10 (or 1+0) is the optimal choice for write-intensive database applications. For every write operation, the data is written twice (once to each disk in a mirrored pair), resulting in a write penalty of 2. In contrast, RAID 5 requires a read-modify-write sequence for each write (read data, read parity, write new data, write new parity), resulting in a write penalty of 4. RAID 6 is even worse with a write penalty of 6. Given the requirement for the lowest write latency, RAID 10's lack of parity calculation overhead makes it superior to both RAID 5 and RAID 6, despite having lower usable capacity (4TB vs. 7TB for RAID 5 or 6TB for RAID 6).

This is a classic virtualization performance problem. The guest OS reports low CPU usage (% Processor Time) because it is not actually getting scheduled onto a physical CPU core. The high Processor Queue Length inside the VM indicates that threads are ready to run but cannot. This discrepancy occurs when the hypervisor host is CPU-constrained and cannot service the VM's requests for CPU time in a timely manner. This wait state is often measured by a hypervisor-level metric called 'CPU Ready' or 'CPU Steal'. The other options are contradicted by the provided counters: low disk queue negates a storage bottleneck, high available memory negates a memory issue, and while a network issue is possible, the direct evidence points strongly to CPU scheduling contention on the host.

This scenario highlights a critical disaster recovery mistake. Restoring DHCP-A from a backup taken before the failover configuration means it has no knowledge of the partnership with DHCP-B. When it comes online, it will believe it is the sole authoritative DHCP server for the scope and will begin issuing leases from its old, outdated database. Since DHCP-B is also actively managing the scope, this will inevitably lead to IP address conflicts on the network. The correct recovery procedure involves rebuilding the server and re-establishing the failover partnership from the active server (DHCP-B), not restoring an old backup.

The key constraints are leveraging existing 10GbE infrastructure and the lack of Data Center Bridging (DCB) support. Fibre Channel is eliminated as it needs a separate, dedicated network. FCoE is a strong contender for running block storage over Ethernet, but its requirement for a lossless network, typically provided by DCB, makes it unsuitable for the existing infrastructure. NFS is file-level, not block-level. iSCSI is the perfect fit because it is designed to run over standard TCP/IP Ethernet networks, provides the required block-level access, and does not depend on DCB, making it the most cost-effective and technically compatible solution.

This question requires calculating two separate metrics. First, calculate the total downtime against the SLO. A 30-day month has minutes. The allowed downtime for 99.9% uptime is minutes. The total actual downtime was minutes. Since , the uptime SLO was met. Second, evaluate the MTTR. MTTR is the average time to recover from failures. The average recovery time for the two outages is minutes. Since $16$ minutes is greater than the 15-minute MTTR defined in the SLA, the MTTR was breached. Therefore, the uptime SLO was met, but the MTTR was breached.

The error code 0x80244019 corresponds to an HTTP 404 'Not Found' error. In a WSUS context, this means the client successfully contacted the WSUS server and received the list of approved updates (metadata), but when it tried to download the actual update files from the content directory, the files were not present. Since other branches are working, this points to an issue with the specific downstream replica server. The most common cause for this is that the metadata (the approval for the update) has replicated from the upstream server, but the much larger update binaries have not yet been downloaded and stored locally on the downstream server's content store. The other options are less likely: a corrupted agent or BITS issue would likely affect all updates, not just a new one, and a firewall issue would prevent contact entirely, not cause a 404 error.

This question tests deep knowledge of JEA configuration. The goal is to constrain a standard cmdlet (Restart-Service) to a specific parameter value. Option A is too permissive; it allows restarting any service. Option C relies on a custom function which is a valid approach, but not the most direct way to constrain a built-in cmdlet. Option D uses an alias, but aliases in JEA do not provide a security boundary; the user could still call Restart-Service directly with other parameters. Option B is the correct and most secure method. It makes the Restart-Service cmdlet visible but constrains its -Name parameter to a ValidateSet containing only 'SQLSERVERAGENT'. This prevents the user from restarting any other service while still using the standard, familiar cmdlet.

Effective capacity planning requires not just looking at long-term averages but also identifying and weighting recent trends. The long-term average of 2% per month would suggest the 15% buffer (from 30% down to 15%) would be consumed in months. However, the recent, more relevant data shows a new usage pattern of 4% per month. This is a more accurate predictor of future behavior. Using this rate, the 15% buffer will be consumed in months. Therefore, scheduling the upgrade in approximately 4 months is the most prudent action based on a proper analysis of the changing trend.

This question probes the specific behavior of NIC Teaming modes. In Switch Independent mode, the server's teaming software makes load balancing decisions without coordination from the switch. The Dynamic algorithm (or Address Hash in older versions) balances outbound traffic by distributing different TCP streams (conversations) across the available NICs. However, to prevent out-of-order packet delivery which would cripple TCP performance, it ensures that all packets for a single conversation (defined by source/destination IP/port) always exit through the same physical NIC. Therefore, the aggregate throughput can exceed one NIC's capacity by serving multiple clients, but any single client's stream is limited to the bandwidth of the one NIC it's been assigned to (~10 Gbps).

This is a complex interaction between server roles. When the Hyper-V role is installed, the server's architecture fundamentally changes. The hypervisor (a Type-1, bare-metal hypervisor) loads first and takes direct control of the hardware. The original Windows Server OS is then loaded into a special virtual machine called the parent partition (or root partition). While the parent partition has privileged access, the hypervisor still abstracts and controls direct hardware access. A service that expects to read low-level CPU registers directly will fail because the hypervisor is now intercepting these calls. This is a key reason why it's a best practice not to install other roles, especially those with low-level hardware interaction, on a Hyper-V host. While running a DC on a Hyper-V host is also not recommended for other reasons, it is not technically impossible; the direct cause of the failure is the hardware abstraction layer introduced by the hypervisor.

This scenario describes a common challenge with thin provisioning. When a file is deleted inside a guest OS, it simply marks the blocks as free in its own file system table; it does not typically inform the underlying storage that these blocks are no longer needed. From the SAN's perspective, the blocks are still allocated. To reclaim this space, a command must be issued that tells the storage array which blocks are free. This is achieved through the SCSI UNMAP command (or ATA TRIM command for SATA/NVMe). Most modern hypervisors and guest OSes support this, but it often needs to be enabled or run periodically as a scheduled task to 'punch zeros' or send UNMAP commands for the freed blocks, allowing the thin-provisioned LUN to shrink.

This scenario illustrates the infamous 'RAID 5 write hole' or, more accurately, the risk of UREs during a rebuild. In a degraded RAID 5, the data on the failed drive is reconstructed by reading data from all other surviving drives and using the parity information. If a read error occurs on any of these surviving drives during the rebuild, the controller cannot reconstruct the data for that specific stripe. Since RAID 5 has only a single parity block, it cannot recover from a second failure event (the original failed drive + the URE on another drive). This leads to a catastrophic failure of the rebuild process and the loss of the entire array. This high-risk scenario is a primary reason why RAID 6 (with its dual parity) is recommended for large arrays of high-capacity drives.

This question requires a nuanced understanding of application-specific performance monitoring. SQL Server is designed to be memory-intensive. It will intentionally use as much RAM as it is allocated to cache data pages from the database, which minimizes slow disk I/O. Therefore, seeing low Available MBytes on a SQL server is expected. The key confirming counters are Pages/sec (which indicates hard paging to disk) being near zero, and Page life expectancy (how long data pages stay in the cache) being high and stable. These two counters together confirm that SQL Server is using its memory cache effectively and is not under memory pressure. Adding more RAM would likely just result in SQL Server using that new RAM for its cache as well, without necessarily improving performance if the cache is already effective.

The WinSxS folder contains components that allow for Windows features to be added and updates to be uninstalled. Manually deleting files from it will corrupt the operating system. sfc and chkdsk are for system file integrity and disk errors, not cleanup. The Disk Cleanup utility is a valid method, but the DISM command offers more power and control. The command Dism.exe /Online /Cleanup-Image /StartComponentCleanup removes superseded versions of components. Adding the /ResetBase switch is a more aggressive step that removes all superseded versions of every component in the component store, and critically, it makes all existing updates and service packs permanent, meaning they can no longer be uninstalled. In a scenario where space is critical on a stable, long-running server, this is the most effective command to maximize space reclamation from the component store.

This scenario describes the classic 'Kerberos double-hop' problem. Standard Kerberos authentication prevents a service from forwarding a user's credentials to another service to prevent credential theft. To allow this for multi-tier applications, Kerberos Delegation must be configured. The middle-tier server needs to be trusted in Active Directory to impersonate the user and present delegated credentials to the SQL server. Specifically, Kerberos Constrained Delegation (KCD) is the modern, secure method where you configure the middle-tier computer object to be trusted to delegate to specific Service Principal Names (SPNs), such as the MSSQLSvc SPN on the database server. Without this configuration, the second hop (middle-tier to SQL) will fail authentication.

The key to this question is matching the I/O pattern to the storage technology. The workload is large-block sequential reads. Traditional spinning disks (HDDs) excel at this type of workload because the slowness of seek time is minimized; once the read head is in position, it can read long, contiguous streams of data very quickly. A large number of spindles (drives) in a RAID array will provide very high aggregate sequential throughput. An all-flash array would offer superior performance, but its primary benefit is for small-block, random I/O, making it an unnecessarily expensive solution for this specific workload (not cost-effective). A hybrid array's cache would be less effective for scanning massive datasets that are much larger than the cache size. Object storage is not suitable for the performance demands of a data warehouse query engine.

This is a common capacity planning pitfall. Not all workloads scale with more cores. A single-threaded application can only execute on one CPU core at a time. Therefore, its performance is almost entirely dependent on the clock speed (Instructions Per Clock x Frequency) of that single core. The upgrade from a 3.5 GHz core to a 2.5 GHz core represents a significant decrease in single-threaded performance, even though the new CPU's total aggregate computing power (16 x 2.5 GHz) is much higher. The application simply cannot use the other 15 cores, so the 'upgrade' was actually a downgrade for this specific workload.

This question highlights that SLAs are often multi-faceted. Focusing only on uptime is a mistake. The 95th percentile metric means that 95% of requests must be faster than the threshold. The 3-hour peak period constitutes of the month. However, during that time, all users were experiencing high latency. The vast majority of the month (99.6%) had fast responses. The 95th percentile calculation would look at all response times, and since more than 5% of the time period experienced responses well above 200ms, the 95th percentile value would certainly be higher than 200ms. Therefore, even with 100% uptime, the performance aspect of the SLA was breached.

Write amplification is a critical concept for SSDs. It's the ratio of physical writes to the SSD versus logical writes from the host. RAID levels that use parity, especially with small, random write workloads typical of OLTP, suffer from high write amplification at the array level. For a small write in RAID 5, the controller must read the old data block, read the old parity block, calculate the new parity, write the new data block, and write the new parity block. This turns a single host write into two reads and two writes at the drive level, significantly increasing the total write I/O and wear on the SSDs. RAID 1 and 10 have a simple, fixed write penalty of 2 (the data is just written to two drives). While RAID 0 has the lowest amplification, it's unsuitable due to lack of redundancy. Therefore, RAID 5 is the most damaging choice for SSD endurance in a write-heavy OLTP environment.