Cisco 350-401: ThousandEyes Path Analysis

When an application is slow, the network team rarely owns every hop between the user and the service. The path may cross a campus, ISP, internet exchange, cloud edge, SaaS provider, or remote data center, and a traditional device-by-device view can end at the administrative boundary long before the fault. Cisco ThousandEyes helps by combining end-to-end measurements with hop-by-hop path visualization from distributed agents, but the data is most useful when engineers understand what the probes actually measure and what they do not prove.

The current Cisco 350-401 ENCOR v1.2 Network Assurance domain includes diagnosing network problems with tools such as ping, traceroute, SNMP and syslog; configuring Flexible NetFlow, SPAN, IP SLA and NETCONF/RESTCONF; and describing Cisco Catalyst Center assurance. ThousandEyes is not separately named as a mandatory product objective in that public v1.2 blueprint. It is valuable adjacent engineering practice: ThousandEyes network tests distinguish agent-to-target loss, latency and jitter from the traceroute-like forwarding view in Path Visualization, which helps operators locate evidence before assigning incident ownership or changing infrastructure.

In Cisco Network Engineering, path analysis belongs between the reported symptom and the infrastructure change: evidence should locate where behavior changed before a team edits routing, policy, or transport settings. The goal is to locate where behavior changed, compare that with a known baseline, and identify which team or provider has the evidence needed for the next action.

End-to-end metrics and path visualization are complementary measurements

ThousandEyes documentation distinguishes the end-to-end measurement from the path-visualization measurement. End-to-end tests quantify what happened between the agent and target—loss, latency, jitter, and other metrics depending on test type. Path visualization discovers the intervening hops and displays the observed path. A network can show high end-to-end loss while an individual hop does not expose a clear loss signal, or a hop can appear unusual without causing application impact.

In ThousandEyes monitoring, keep end-to-end loss and latency separate from hop-by-hop path evidence so a degraded application path is not blamed on the first router that happens to appear in a trace. Use the timeline to establish when user-impacting performance changed, then use path visualization to investigate whether route or forwarding behavior changed during the same interval.

The test type also matters. Agent-to-server and agent-to-agent tests can expose different measurements and directions. Agent-to-agent testing can reveal asymmetric behavior because measurements can be taken from either side, while a single-source test may only show one operational perspective.

A hop that does not reply is not automatically a failed hop

Path discovery relies on TTL-expired responses or related probe behavior, and intermediate devices can rate-limit, filter, deprioritize, or suppress those replies. A node that looks blank or shows apparent forwarding loss may still be passing production traffic correctly. The decisive question is whether loss continues through subsequent hops and appears in the end-to-end measurement.

This distinction prevents a common false accusation: identifying the first device that does not answer probes as the outage source. If later hops and the destination continue to respond without equivalent loss, the missing reply is more likely a visibility characteristic than a forwarding failure. Operators should compare downstream continuity before escalating to a provider.

Device counters and network-state monitoring describe the assets an organization administers, while ThousandEyes tests can expose service-path behavior across third-party networks. A comparison of ThousandEyes and SolarWinds monitoring is a useful way to understand which questions each product answers, but no one monitoring platform automatically provides complete end-to-end root cause. The destination here is specifically about ThousandEyes versus SolarWinds—not a generic tutorial for every monitoring product.

TCP and ICMP path discovery can reveal different network behavior

Current ThousandEyes documentation explains that path visualization may use ICMP echo requests or TCP SYN packets depending on the test protocol. Firewalls, ACLs, middleboxes, and provider policies can treat those probe types differently. A path discovered with ICMP is not necessarily identical to the forwarding behavior of the application’s TCP flow under every network policy.

For an HTTPS service, a TCP-oriented test can better approximate whether the network permits traffic toward the relevant destination and port, while ICMP may be filtered even when the application is healthy. Conversely, TCP behavior can be influenced by service or security devices in ways that make ICMP useful for a simpler Layer 3 comparison. The best protocol is the one aligned with the diagnostic question.

Engineers should also document proxy behavior. Endpoint agents can send web application traffic through a proxy while network tests observe the path to that proxy rather than the final SaaS destination. If the troubleshooting team does not know that, it can spend hours investigating the wrong path.

Route changes matter most when they correlate with a user-impacting change

Internet paths change constantly. A new autonomous system, different transit provider, or alternate ingress point is not automatically a problem. The useful question is whether the route change aligns with the start of higher latency, packet loss, jitter, or application failure. Correlation turns path visualization from an interesting map into operational evidence.

A baseline should record common paths by location and time period. If one branch office normally reaches a SaaS edge through a particular provider and suddenly shifts through a longer path during the incident window, the change deserves investigation. If the application remained healthy, the route change may simply be normal internet adaptation.

Network and application observations should be correlated to distinguish transport delay from slow server-side processing. A comparison of ThousandEyes and Dynatrace observability helps clarify the different vantage points, but this destination specifically contrasts ThousandEyes and Dynatrace rather than establishing that one universal observability architecture will fit every incident. Healthy network tests do not exclude an application fault; a slow transaction alone does not prove the path failed.

Directional analysis is critical when paths are asymmetric

IP routing is often asymmetric: the path from user to service differs from the path from service to user. A one-direction investigation can therefore miss the problem. Agent-to-agent tests that measure both directions are valuable because they can expose different path branches, latency, or loss depending on direction.

Asymmetry is not inherently wrong, but it matters when firewalls, stateful inspection, QoS, or provider capacity differ between directions. A user may experience poor download performance while the upload path remains normal. Treating one round-trip number as proof that both directions are healthy can hide that distinction.

The investigation should note where agents are placed and which traffic direction is being observed. Without that context, two screenshots from the same incident can appear contradictory when they are actually measuring different directions or test sources.

DSCP observations can expose QoS boundary problems

ThousandEyes network tests can set DSCP on probe packets, and path visualization can expose DSCP changes where supported. That is useful when a high-priority application behaves correctly inside the enterprise but degrades after crossing a boundary that remarks or clears the marking. The observation does not by itself prove queue treatment, but it can identify where the packet’s requested service class changed.

DSCP marking should be interpreted alongside device QoS policy and congestion evidence. A packet can retain its DSCP value while receiving poor service because the downstream network does not honor it; another network can remark traffic intentionally as part of an agreed trust boundary.

This is why end-to-end QoS troubleshooting needs both packet marking and performance measurement. Marking tells the network what class the traffic requests. Latency, jitter, loss, and queue counters reveal what happened under congestion.

Use path evidence to improve escalation quality

External-provider incidents are resolved faster when the escalation contains time, source agent, destination, affected path, hop or link behavior, loss and latency evidence, and a comparison with a healthy period. “The internet is slow” gives the provider little to investigate. A bounded path change with timestamps and multiple affected agents gives them a defensible starting point.

Endpoint Agents can add the actual user location to that evidence, especially for remote work. Cloud agents and enterprise agents observe infrastructure vantage points; endpoint agents can reveal how the path looks from the user’s device and access network. The correct vantage point depends on whose experience the incident is trying to explain.

The broader Cisco assurance strategy benefits when ThousandEyes data is correlated with routing, interface, controller, and application signals rather than used as a standalone verdict. Path analysis is most powerful when it narrows the failure domain and tells the next owner what to inspect.

A repeatable path-analysis workflow prevents visualization from becoming guesswork

Start with the user symptom and time window. Confirm the end-to-end metric that changed. Compare affected and unaffected agents. Inspect path visualization for branches or hops that changed at the same time. Check direction and test protocol. Correlate with routing, interface, QoS, DNS, proxy, or application telemetry. Only then identify the likely administrative domain responsible for the next investigation.

Preserve the healthy baseline as well as the incident data. A screenshot after the provider has recovered can look perfectly normal and erase the evidence that made the incident diagnosable. Scheduled tests provide the historical context needed to compare before, during, and after the event.

ThousandEyes path analysis is strongest when each result is interpreted within its measurement limits. Agent-to-target data indicates tested loss, latency or jitter; Path Visualization shows observed responses and forwarding-path structure, not an omniscient reconstruction of all production packets. Reverse-direction measurements, agent placement and correlated device/application telemetry can help identify a failure domain, but path visualizations alone cannot prove the source of every application problem.

BGP path views add control-plane evidence to forwarding observations

For internet-facing destinations, ThousandEyes can supplement hop-by-hop forwarding data with BGP visibility when the test and monitors are configured for it. That helps distinguish a local forwarding problem from an external routing change that altered how a prefix is reached across autonomous systems. The two views should be correlated rather than treated as interchangeable.

A BGP announcement change may explain why traffic moved to a different provider even though no enterprise router changed configuration. Conversely, the BGP path can remain stable while a lower-level forwarding path inside one AS changes and creates latency. Engineers should therefore record the control-plane change, the forwarding-path change, and the user metric that changed at the same time.

This combination improves provider escalations because it shows whether the incident began with reachability, path quality, or routing advertisement behavior. It also prevents overfitting to one visualization: no single layer describes the entire journey from enterprise agent to application.

Leave a Reply

How It Works

img
Step 1. Choose Exam
on ExamLabs
Download IT Exams Questions & Answers
img
Step 2. Open Exam with
Avanset Exam Simulator
Press here to download VCE Exam Simulator that simulates real exam environment
img
Step 3. Study
& Pass
IT Exams Anywhere, Anytime!