Prepare and test your skills

Prepare and test your skills

Free sample question with answers and detailed explanation below. The Professional Cloud Network Engineer practice bank has an additional 2,437 practice questions.Try 10 free realistic questions: instant score, no credit card, and no account required.
You are troubleshooting a 15-minute performance degradation between Compute Engine VMs in us-central1-a and europe-west1-b within a single VPC network. In Network Intelligence Center's Performance Dashboard, you observe the following for that zone pair:
Your us-central1-a deployment includes preemptible VMs and a third-party VPN gateway running a custom userspace TCP/IP stack.
How should you evaluate the scope and source of these metrics?
Comparing the Per Project view against the Google Cloud view in Performance Dashboard is the primary method for isolating whether network degradation stems from Google's global backbone or project-specific resources. Because the Google Cloud view for us-central1-a to europe-west1-b remained completely healthy (near 0% loss and steady 102 ms latency), the underlying physical network path is functioning normally and the anomalies are scoped to your project.
us-central1-a and europe-west1-b.*).SEQ to ACK intervals) rather than ICMP. Applications with non-standard or userspace TCP/IP stacks—such as third-party VPN applications—introduce processing delays into TCP acknowledgments, which Performance Dashboard records as elevated network RTT.*) simply confirms sufficient VM density per zone (>= 2.5 * zones), making the project metric statistically valid for your VM mesh while still reflecting VM-level events.This assessment accurately correlates the probing methodologies of both metrics with the divergence between Google-wide and project-scoped baselines, preventing unnecessary escalation for a backbone issue and directing troubleshooting toward preemption events and VPN VM TCP processing.
In Performance Dashboard, latency (RTT) is never measured using active synthetic probing in either the Per Project or Google Cloud view. Both scopes rely on passive probing of actual customer TCP traffic headers—using the same data source and sampling methodology as VPC Flow Logs—to calculate the elapsed time between a TCP sequence number (SEQ) and its corresponding acknowledgment (ACK).
ACK is observed, delays inside a VM's custom TCP/IP stack or slow application-level responses can artificially inflate reported latency.To measure packet loss without consuming guest CPU or appearing as traffic inside guest network interfaces, Performance Dashboard runs probing workers directly on the physical hypervisor hosts that house Compute Engine VMs. These workers inject and receive synthetic UDP probes (for internal IPs) and TCP probes (for external IPs) across the project's reachable VM mesh.
us-central1-a and europe-west1-b, the Google Cloud performance view for that identical zone pair would also reflect elevated packet loss rather than remaining near 0%.us-central1-a and europe-west1-b, while Google Cloud median latency remained at 102 ms and only your project's latency rose to 240 ms due to the VPN appliance's TCP stack.us-central1-a and europe-west1-b that is masked by cross-project averaging in the Google Cloud view, and the absence of an asterisk confirms physical network packet loss.The Google Cloud performance view in Performance Dashboard aggregates packet loss and median latency (RTT) metrics across all Google Cloud projects on a per-zone-pair basis. It serves as an infrastructure baseline to determine whether a network issue is affecting Google's underlying backbone between two specific zones or is unique to a single project.
us-central1-a to europe-west1-b path across all customer traffic and active probes, a physical backbone fault between those two zones would appear as elevated loss or latency in the global view rather than being hidden by unrelated regions.*) in the Per Project packet loss heatmap indicates only that the project had enough VMs per zone to meet the 90% or 95% statistical confidence threshold (at least 2.5 or 10 VMs multiplied by the number of zones, respectively). It does not filter out project-specific VM lifecycle events.us-central1-a to europe-west1-b zone pair during the same 15-minute window, confirming a shared physical backbone incident.Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated Validated