Cisco 300-445 Designing and Implementing Enterprise Network Assurance(300-445 ENNA) Exam Practice Test

As shown inExhibit 4.1-1(image_6e19fb.png), the alert rule is configured for the BGP layer with the name "AS origin AS 1234". The critical logic is defined in theAlert Conditionssection, which specifies: BGP Origin meets all of: ASN not in 1234.

This specific configuration is the standard method used in ThousandEyes to detectBGP hijacking(Option C). BGP hijacking occurs when an unauthorized Autonomous System (AS) announces a network prefix that it does not own, effectively stealing the traffic meant for the legitimate owner. By setting an alert that triggers when a monitored prefix is seen originating from an Autonomous System Number (ASN) that isnotthe expected authorized ASN (in this case, AS 1234), ThousandEyes provides immediate notification of a routing takeover.

Identifying this via BGP Origin alerts is superior to simple connectivity checks because it catches the issue at the routing control plane.

DNS Issues (Options B and D):These involve the translation of names to IPs and would be detected via DNS-specific tests rather than BGP Origin checks.

Route Poisoning (Option A):This is a generic term often referring to the manipulation of distance-vector protocol metrics to make a route unreachable; while related to routing, BGP Origin checks are the specific tool for detecting external prefix hijacks.

By monitoring the ASN origin, network engineers can quickly distinguish between a local ISP outage and a global security event where their traffic is being misdirected to an illegitimate network.

Troubleshooting intermittent performance issues is a core component of theDesigning and Implementing Enterprise Network Assurance (300-445 ENNA)curriculum. When a transaction test using Basic Authentication fails randomly, the engineer must employ a multi-layered diagnostic approach to determine if the fault lies with the credentials, the network path, or the application server.

The first step is toverify the correctness of credentials(Option B). Because Basic Authentication encodes credentials in the header, any change in user permissions or account lockouts will trigger immediate failures. Performing manual logins from various geographical locations helps rule out location-based access control lists (ACLs) or regional IdP sync issues.

The most data-rich diagnostic step is toanalyze the ThousandEyes waterfall charts and HTTP response codes(Option C). The waterfall view allows the engineer to see if the failure happens during the initial 401 Unauthorized challenge or after the credentials are sent. If the charts show high "Wait Time" or 5xx errors, the issue is likely server-side latency or instability. If the "Connect Time" is high, the problem may be network-layer congestion.

Finally, if the telemetry indicates that the network path is healthy but the server is intermittently returning errors or timing out, the engineer shouldcontact the web application vendor(Option D). Providing the vendor with the specific ThousandEyes "Share Link" allows them to see the exact same packet-level and browser-level data, proving that the issue is not with the client's network but with the server-side infrastructure. Disabling authentication (Option A) is not a valid troubleshooting step for a test designed specifically to monitor an authenticated workflow.

In theDesigning and Implementing Enterprise Network Assurance (300-445 ENNA)curriculum, identifying routing anomalies such as BGP Hijacking is a critical aspect of external network assurance. A BGP Hijack occurs when an Autonomous System (AS) illegitimately announces a prefix it does not own, causing traffic intended for the rightful owner to be misdirected.

By contrastingExhibit 3.4-1(Before the Outage) andExhibit 3.4-2(During the Outage), a definitive change in the routing path is observed. In the baseline view (Exhibit 3.4-1), traffic from the Los Angeles agent reaches the destination network 54.239.104.0/23, which is correctly associated withAmazon.com, Inc. (AS 16509). However, during the outage (Exhibit 3.4-2), the path visualization shows the traffic being redirected into a different network:eNET Inc. (AS 10297). At this new location, the traffic encounters100% Forwarding Lossat the node ten7-8.core-2.xlhost.com.

Thechange in the Autonomous System (AS) numberfrom 16509 to 10297 (Option B) is the most concrete indicator of a BGP Hijack. While symptoms such asAvailability Drop(Option A),Response Delay(Option C), andPacket Loss(Option D) are clearly visible in the ThousandEyes telemetry, these symptoms are generic and could be caused by various other issues like physical link failures or congestion. The shift in the AS path—specifically to an AS that does not legitimately host the target's IP prefix—provides the forensic evidence needed to identify the root cause as a routing takeover. This visualization allows network administrators to bypass internal troubleshooting and immediately focus on external remediation with upstream providers.

According to theDesigning and Implementing Enterprise Network Assurance (300-445 ENNA)curriculum, the native integration of ThousandEyes within the Meraki portfolio is specifically focused on theMeraki MXsecurity and SD-WAN appliances. This integration allows users to install ThousandEyes Enterprise Agents directly on supported Meraki MX hardware, providing critical visibility into the performance of services that distributed users rely on.4

TheMeraki MX (Option A)supports the ThousandEyes Enterprise Agent as a containerized service that can be activated with a single click within the Meraki Dashboard.5This integration is designed to provide better monitoring and testing capabilities for customers interested in improving the quality of their experience and adding the appropriate SD-WAN policies to optimize network performance. By running the agent natively on the MX, administrators can monitor the performance of external applications and services directly from the SD-WAN sites, bridging the gap between local branch health and global internet performance.6

While Cisco continues to expand its assurance integrations across the portfolio, theMR (Wireless),MS (Switching),MV (Camera), andMG (Cellular)series do not currently support the native, "on-box" execution of the ThousandEyes Enterprise Agent. For these other platforms, visibility is typically achieved through passive monitoring viaMeraki Insight (MI)or by deploying ThousandEyes agents on connected client devices (Endpoint Agents) or nearby infrastructure. Therefore, theMeraki MXis the only verified platform supporting native ThousandEyes agent deployment in this context.

In theDesigning and Implementing Enterprise Network Assurance (300-445 ENNA)curriculum, the evolution of network monitoring includes moving from periodic polling to real-time data streaming. The exhibit displays a curl command targeting the ThousandEyes API v7 /stream endpoint. When integrating ThousandEyes with high-performance observability platforms likeGrafana, the standardized and recommended method for machine-to-machine data exchange is throughOpenTelemetry (OTel).

According to the ENNA architecture guidelines, the ThousandEyesStreaming APIallows users to push granular test metrics (such as network latency, packet loss, and jitter) to external collectors in an OTel-compatible format. In the provided JSON payload, the "type" field is a mandatory parameter that defines the integration protocol. For Grafana, which natively supports OpenTelemetry Protocol (OTLP) via its OpenTelemetry Collector, the value must be set to"opentelemetry"(Option A). This tells the ThousandEyes streaming engine to encapsulate the data according to the OTel semantic conventions, ensuring that Grafana can correctly interpret and visualize the metrics without additional custom parsing logic.

While other options exist in the ThousandEyes ecosystem, they do not fit the specific API call shown for this use case:

Custom Webhooks(Option B) are typically used for event-driven alerts and notifications (e.g., sending a POST request when a threshold is breached) rather than continuous high-fidelity metric streaming.

Push-apiandpoll-api(Options C and D) are not valid "type" values within the context of the v7 /stream endpoint, as the streaming service specifically utilizes the OpenTelemetry framework for real-time delivery.

By selectingopentelemetry, the network engineer enables a robust "push-based" integration that provides real-time visibility into application performance and network health, leveraging Grafana's advanced dashboarding capabilities to analyze ThousandEyes telemetry data alongside other enterprise infrastructure metrics.

Introduction to ThousandEyes for OpenTelemetry

This video provides a foundational understanding of how ThousandEyes uses modern streaming frameworks to export critical performance data to external observability platforms.

According to theDesigning and Implementing Enterprise Network Assurance (300-445 ENNA)curriculum,ThousandEyes WAN Insightsis a predictive analytics solution designed to transform network management from a reactive to a proactive model.10In the scenario provided, the voice quality issues are recurring and peak-hour dependent, indicating a need for optimization rather than just standard real-time alerting.11

The most relevant capability for this scenario isanalyzing historical SD-WAN performance datato generatepath recommendations(Option B). WAN Insights integrates withCisco Catalyst SD-WAN Manager(vManage) andvAnalyticsto ingest massive volumes of telemetry.12It uses advanced statistical models to analyze the performance of all active circuits (MPLS, Internet, etc.) over time. If the system identifies that a different available path consistently delivers a higher Quality of Experience (QoE) or better adherence to the voice SLA during those peak windows, it generates a recommendation to adjust theApplication-Aware Routing (AAR)policy.1314

This predictive approach allows the administrator to fine-tune network policies in advance, ensuring that sensitive traffic like voice is automatically rerouted to the most stable path before the congestion impacts the users. Option A describes standard ThousandEyes synthetic testing, while Option C is incorrect because WAN Insights specifically uses existing SD-WAN telemetry rather tha15n generating its own synthetic voice16probes. Option D and E describe general NMS or security features not specific to WAN Insights' predictive mission. Thus, the proactive recommendation of alternative paths based on historical SLA adherence is the core function of WAN Insights for improving voice quality.171819

In theDesigning and Implementing Enterprise Network Assurance (300-445 ENNA)framework, baselining is the process of establishing a "normal" performance profile for network infrastructure to enable the detection of anomalies. When the metric of interest is the internal health of a physical device, such asCPU utilizationon a core router,SNMP (Simple Network Management Protocol)is the industry-standard data source.

SNMP provides direct visibility into the device's control plane and hardware performance. By polling specific Object Identifiers (OIDs) from the router's Management Information Base (MIB), a monitoring system likeCisco Catalyst Centeror a third-party NMS can collect granular data on CPU cycles, memory allocation, and temperature. This "inside-out" telemetry is essential for baselining because it reflects the actual resource consumption of the router during various traffic loads.

Conversely, ThousandEyes tests (Options A, B, and D) provide "outside-in" synthetic data. WhileDNS resolution time(Option A),HTTP response times(Option B), andPath Visualization(Option D) are excellent for measuring end-to-end service delivery and network transit health, they do not report on the router's internal hardware state. For instance, a router could have 99% CPU utilization (indicating a potential crash), yet a ThousandEyes path test might still show a "green" path if the data plane (ASICs) is still forwarding packets efficiently. Therefore, to establish a reliable baseline for hardware-specific metrics like CPU,SNMP data(Option C) is the only appropriate source among the choices.

In theDesigning and Implementing Enterprise Network Assurance (300-445 ENNA)curriculum, transaction monitoring is essential for validating complex, multi-step user workflows. When an engineer with limited scripting experience needs to monitor a OneDrive file download process, the ThousandEyes ecosystem provides several tools to simplify the creation of robustTransaction tests.

The correct approach is to leverage all available resources, makingOption Dthe definitive answer. First, theThousandEyes Recorder IDE(Option B) is a critical tool for non-scripters. It allows an engineer to perform the actual workflow—navigating to OneDrive, logging in, and initiating the download—in a browser environment on their local machine while the tool records every click and keyboard entry. The Recorder automatically generates the corresponding JavaScript code using theThousandEyes transaction library. Second, the platform provides pre-builtscript templates(Option A), such as those forOffice 365andOneDrive, which include baseline logic and best practices for these specific services. Third, the officialtransaction-scripting-examplesrepository on GitHub (Option C) is a maintained source of code snippets. This is particularly useful for implementing advanced logic, such as aretry mechanism, which ensures that the test does not report a "failure" due to a transient network hiccup but instead attempts the action again before triggering an alert.

As shown in the provided exhibit, a typical script uses import statements from @thousandeyes and selenium-webdriver to control the browser. By combining theRecorderfor the basic flow, aTemplatefor service-specific nuances, andSample Scriptsfor logic enhancements like retries, the engineer can deploy a highly reliable assurance test without deep coding expertise. Therefore, all three actions are highly recommended and valid within the ENNA implementation framework.

56

In theDesigning and Implement7ing Enterpr8ise Network Assurance (300-445 ENNA)framework, configuring effective alert rules is critical for distinguishing between standard network noise and actionable application-layer failures. ForWeb - HTTP Servertests, ThousandEyes allows engineers to monitor both network-level metrics (like Connect time) and application-level indicators (like HTTP response codes).

The requirement is to trigger an alert specifically when the HTTP server responds with aserver error. In the HTTP protocol, server errors are categorized as the5XXseries of status codes (e.g., 500 Internal Server Error, 503 Service Unavailable, 504 Gateway Timeout). To meet this requirement, the engineer must configure a location alert condition where theMetricis set toResponse Codeand the condition value isserver error(5XX)(Option D).

Reviewing the other options:

Error type is any (Option A):While this would capture server errors, it would also trigger for 4XX client errors (like 404 Not Found) and network-layer timeouts, making it too broad for a specific "server error" requirement.

Wait Time is Dynamic (Option B):This monitors the time-to-first-byte using statistical baselining. While high wait times oftenprecede5XX errors, this condition only alerts on latency, not on the actual error code itself.

Response Time (Option C):Similar to wait time, this monitors performance speed rather than the logical success or failure of the server's response.

By specifically selectingResponse Code: server error(5XX), the engineer ensures that the operations team is only notified when the application backend is experiencing a functional failure, rather than just a slow response or a client-side misconfiguration.

Establishing an accurate baseline is a cornerstone of theDesigning and Implementing Enterprise Network Assurance (300-445 ENNA)methodology. A baseline must represent the "typical" behavior of the network across its standard operating cycle to be useful for anomaly detection and capacity planning. When monitoring interface utilization on a critical link, it is vital tocapture both peak and off-peak traffic patterns(Option C).

Most enterprise networks exhibit cyclical traffic patterns, often referred to as "diurnal" cycles, where utilization spikes during business hours (peak) and drops significantly at night or on weekends (off-peak). If an engineer only collects data during the lowest volume period (Option A), the baseline will be unrealistically low, leading to "false positive" alerts when normal business traffic returns. Conversely, only capturing peak data might hide background synchronization tasks or backup windows that occur during off-hours.

By selecting a representative time period—typically one full week or a month—the engineer ensures the baseline accounts for variations such as Monday morning "logon storms," mid-week steady states, and weekend maintenance. This comprehensive view allows the assurance platform (such asCisco Catalyst CenterorThousandEyes) to calculate a standard deviation and establish dynamic thresholds. If utilization deviates from these historically representative norms, the system can trigger an alert based on a true anomaly rather than a predictable daily surge. Aligning with a financial year (Option B) is irrelevant to technical performance, and limiting the data (Option D) compromises the baseline's statistical validity.

InDesigning and Implementing Enterprise Network Assurance (300-445 ENNA), dashboard widgets must be mapped to the correct telemetry data source and statistical measure. When an agent is installed as aLinux packageon an organization's infrastructure, it is classified as anEnterprise Agent.

The correct configuration for this widget isCloud & Enterprise Agents and Mean(Option B).

Data Source:Since the Linux-based agent is an Enterprise Agent, it shares the same data source category as Cloud Agents in the ThousandEyes dashboard configuration menu.

Measure:To display the "average" packet loss as requested, theMean(the arithmetic average) is the appropriate statistical measure.

Other options are unsuitable:

Option A:Endpoint Agentsare those installed on Windows/macOS user devices, not typically a standard Linux server package used for infrastructure monitoring.Medianwould show the middle value, not the average.

Option C:Routingrefers to BGP data, which does not report packet loss in the same way as synthetic network tests.

Option D:Devicesrefers to hardware monitored via SNMP, and annth Percentileis a specific statistical cutoff (like 95th percentile) rather than a simple average.

For theDesigning and Implementing Enterprise Network Assurance (300-445 ENNA)exam, monitoring remote workforces requires a strategy that captures both the user's local environment and the regional internet health. In a scenario involving Cisco AnyConnect VPN, the "last mile" connectivity of the employee is often the most significant variable in application performance.

Utilizing the ThousandEyes Endpoint Agent(Option D) is the most effective way to monitor this environment. Because the agent resides directly on the remote employee's machine, it can monitor the performance of the web application both "inside" and "outside" the VPN tunnel. It provides visibility into the local Wi-Fi signal strength, the health of the AnyConnect client, and the latency experienced as traffic traverses the VPN headend. This allows engineers to differentiate between a slow home internet connection and an issue with the VPN concentrator.

Deploying ThousandEyes Cloud Agents(Option A) serves as a critical baseline. By running tests from Cloud Agents in the same regions as the remote employees, the engineer can determine if the "internet" in that region is healthy. If a Cloud Agent in London shows a perfect response time while an Endpoint Agent in London shows high latency, the engineer can immediately isolate the problem to the user's specific setup or the VPN path, rather than a regional ISP outage.

Other options are less suitable for monitoring theremote employee's experience:

AppDynamics (Option B)provides server-side code visibility but cannot see the user's home Wi-Fi or local network path.

Enterprise Agents on the VPN concentrator (Option C)can monitor the pathfromthe data center to the app, but they cannot see the path from theuserto the concentrator.

Enterprise Agents in the data center (Option E)provide an "inside-out" view of the app's health but miss the entire remote access experience.

According to theDesigning and Implementing Enterprise Network Assurance (300-445 ENNA)best practices for collaboration monitoring,Microsoft Teams(as well as Webex and Zoom) is monitored using a combination ofScheduled and Dynamic tests(Option D).

Scheduled Tests:These are used to proactively monitor the reachability and performance of the Microsoft Teams infrastructure. The Teams template typically includes scheduled HTTP Server tests to teams.microsoft.com and login.microsoftonline.com, along with network-layer tests to theTeams Transport Relay(e.g., worldaz.tr.teams.microsoft.com). These provide a continuous baseline of connectivity regardless of whether a user is currently in a call.

Dynamic Tests:These are a unique feature for collaboration monitoring. When an Endpoint Agent detects that the Teams application has initiated a real-time media session (audio or video call), it automatically triggers aDynamic Testto the specific IP and port being used for that call. This allows IT teams to see the exact network path—including hop-by-hop latency and loss—of the actual call traffic while the session is active.

Real User Tests (Option C)are generally not used for Teams monitoring in the same way they are for web-based SaaS apps. Because most users utilize the Teamsdesktop clientrather than the browser, the RUM browser extension cannot capture the rich telemetry of the desktop application's internal signaling and media streams. Therefore, the specialized "Automated Session Testing" provided byDynamic Tests, combined with the proactive health checks ofScheduled Tests, constitutes the complete assurance strategy for Microsoft Teams.

In wireless network assurance, high channel utilization and increased retransmissions are clear indicators of RF interference or over-subscription. Retransmissions are particularly damaging to voice call quality because they introduce significant jitter and late-arrival packet loss that the jitter buffer cannot overcome.

The recommended optimization is tochange the access point to a different, less congested channel(Option B). This directly reduces the competition for airtime (Channel Utilization) and the likelihood of collisions from neighboring access points (Co-Channel Interference), which in turn lowers the retransmission rate. Modern wireless controllers usingRadio Resource Management (RRM)often automate this, but a manual adjustment based on SNMP telemetry is a valid operational fix.

Other options are technically flawed:

Option A:Increasing transmit power oftenincreasesinterference and channel utilization for surrounding cells, making the problem worse for everyone.

Option C:Disabling all non-voice traffic is an extreme measure that is rarely feasible in a real business environment.

Option D:Admission control limits thenumberof users but doesn't solve the underlying issue of poor channel health for those already connected.

In the framework ofDesigning and Implementing Enterprise Network Assurance (300-445 ENNA), selecting the appropriate agent for specific test types is a critical configuration step. The ThousandEyes platform utilizes different agent binaries and capabilities depending on the underlying hardware and software environment. Specifically,Page Load testsandTransaction testsrequire the agent to have a browser engine (Chromium) installed and functional to render web content and measure DOM-related metrics.

According to the ENNA implementation guidelines, not all Enterprise Agents support these browser-based tests. The exhibit displays a "New Test" configuration screen where the engineer is searching for "east1-agent-1". While other agents like "west-2-ag-1" are visible, the designated agent is missing because itdoes not support Page load tests(Option D). This common occurrence happens when an Enterprise Agent is deployed on resource-constrained hardware—such as certainCisco Catalyst 9000 switchesorCisco ISR 4000 series routers—where the Chromium browser package is either not supported or has not been enabled due to memory and CPU limitations.

When an agent lacks the required capabilities for a specific test layer (e.g., Web layer tests), it is automatically filtered out from the selection list in the ThousandEyes portal to prevent invalid test configurations. To identify compatible agents, an engineer can look for the "Browser" or "Dual Chromium" labels in the agent settings. If the "east1-agent-1" were merely offline (Option A), disabled (Option B), or still registering (Option C), it would typically still appear in the list but with a status indicator showing its unavailability, rather than being entirely hidden from the test assignment menu. Therefore, the absence of the agent in a web-layer test configuration explicitly indicates a lack of functional support for the required browser-based metrics.

TheDesigning and Implementing Enterprise Network Assurance (300-445 ENNA)framework emphasizes that the goal of internet intelligence is to enable rapid and accurate escalation to the party responsible for a service degradation. Based on the evidence of a BGP Hijack identified in the previous question, the issue is occurring entirely within the public internet ecosystem.

The most appropriate next step is toreach out to the Internet Service Provider (ISP) to report the suspected BGP hijacking incident(Option B). Since the traffic is being misdirected by an external Autonomous System (AS 10297) before it reaches the intended destination (AS 16509), the fix must occur at the routing policy level of the major transit providers. The network administrator should provide the ISP with the ThousandEyes "Share Link" or screenshots showing the path change and the unauthorized AS announcement, as this data serves as proof to accelerate the ISP's mitigation efforts, such as implementing prefix filters or contacting the offending network.

Other options are ineffective for this specific scenario:

Option A:The path visualization shows that traffic is successfully leaving the local network and reaching the public internet; the problem is many hops away from the internal routers.

Option C:Blocking trafficfromAS 10297 does not solve the problem of your traffic beingattractedto it. The hijack affects how the rest of the world (including your ISP) sees the route to your destination.

Option D:DNS is not the issue; the agent successfully resolved the hostname to the correct IP, but the BGP layer misdirected the packets at the routing level.

By identifying the issue as an external routing event, the administrator avoids wasting internal resources and directly triggers the necessary external remediation.

AppDynamics

Catalyst Center

ThousandEyes

Meraki

In theDesigning and Implementing Enterprise Network Assurance (300-445 ENNA)architecture, each platform is positioned to address a distinct domain of visibility and management within the modern IT visibility framework.

AppDynamics(matched to Case 1) is specifically engineered for Full-Stack Observability and Application Performance Monitoring (APM). Its unique focus is linking technical application performance—such as code-level execution and database queries—directly to business outcomes and real-time business metrics. By doing so, it allows organizations to pinpoint how application latency impacts revenue or user satisfaction, making it the primary choice for business impact analysis.

Catalyst Center, formerly DNA Center (matched to Case 2), serves as the foundational controller for campus and enterprise network infrastructure. It leverages AI-driven insights to automate network operations, enforce security policies, and provide at-a-glance health monitoring for local devices and clients. It is the definitive management system for connecting and securing "inside" the corporate network perimeter.

ThousandEyes(matched to Case 3) provides "Internet Intelligence" and is the key solution for monitoring the service delivery chain across environments the enterprise does not own. By utilizing a global network of Cloud, Enterprise, and Endpoint agents, it provides end-to-end visibility from any user location to any application (SaaS or Cloud) over any network (Internet, ISP, or WAN).

Meraki(matched to Case 4) is a cloud-first platform designed for managing highly distributed network environments and IoT deployments. It provides end-to-end visibility through a simplified dashboard that integrates wireless health, switching, and security across thousands of sites, making it ideal for lean IT teams managing distributed retail or branch locations.

Each platform utilizes different data collection methods—active synthetic testing for ThousandEyes and passive monitoring/AI telemetry for Catalyst Center—to ensure comprehensive network assurance across the entire enterprise ecosystem.

Detailed Explanation:

This analysis requires extracting two specific browser-rendering metrics from the IT Operations Dashboard. By placing the cursor over the latest metrics data points:

Page Load time:1479.6 ms.

DOM load time: 1354.3 ms.

By subtracting the DOM time from the Page Load time ($1479.6\text{ ms} - 1354.3\text{ ms} = 125.3\text{ ms}$), the engineer identifies the duration spent rendering the page after the initial document structure was established.

In the context ofDesigning and Implementing Enterprise Network Assurance (300-445 ENNA), analyzing page-load metrics within Cisco ThousandEyes requires identifying the primary bottlenecks that contribute to theTotal Page Load Time. The provided screenshot displays a "Page Breakdown" of 7 resources totaling 953 kB. A critical observation of the pie chart reveals thatImages(the teal-colored segment) constitute the vast majority of the page's payload and resource count.

When the goal is to reduce the page-load time from1023 msto below500 ms, the engineer must target the heaviest components.Lazy loadingis a design pattern that defers the initialization of non-critical resources at page load time. Instead of loading all images simultaneously when the userfirst navigates to the URL, lazy loading ensures that images are only downloaded as they are about to enter the viewport. This significantly reduces the initial DOM load time and the total Page Load Time because the browser does not have to wait for large image files to be fully retrieved before declaring the page "loaded."

Alternative options are less effective in this specific scenario based on the data:

AJAX (XHR/Fetch):The chart shows that XHR and Fetch resources represent a negligible sliver of the total weight; optimizing them would yield minimal gains.

Moving IMG elements:While moving scripts to the bottom can help with rendering, moving image elements to the bottom of the body does not stop the browser from initiating the download requests immediately, thus failing to significantly reduce the total load time.

CDN for Fonts:The "Font" category is also a small fraction of the total 953 kB. While a CDN is a best practice for latency, it does not address the primary "weight" issue caused by the images.

Therefore, implementinglazy loading(Option C) is the most impactful recommendation. It directly addresses the largest resource consumer (Images) identified in the ThousandEyes Page Breakdown, allowing the engineer to reach the sub-500 ms performance target.