Sample questions
Pf Apm Profiling Bottleneck CorrelationDifficulty 1
During a load test, why is it useful to run an APM (Application Performance Monitoring) tool alongside the load generator?
- aAPM replaces the need for a load generator entirely, since some APM agents can also inject synthetic user traffic on their own
- bAPM only matters after the test finishes, to generate the final PDF report
- cAPM captures server-side signals (CPU, GC, DB pool, thread state) at the same moments the load tool records latency.✓
- dAPM is only relevant for functional tests, not performance tests
Explanation:APM tools give visibility into what is happening inside the system under test (CPU, garbage collection, connection pools) at the same moments the load tool measures latency/throughput from outside, which is what enables correlating a client-side symptom with a server-side cause.
Pf Apm Profiling Bottleneck CorrelationDifficulty 1
What is a "bottleneck" in the context of load testing a system?
- aThe specific resource whose saturation caps overall system throughput.✓
- bAny component that consumes CPU during the test, regardless of whether it limits throughput
- cThe virtual user script step that takes the longest to write during scripting
- dThe last request executed before the load test ends
Explanation:A bottleneck is the constraining resource (e.g. DB connection pool, CPU, a single-threaded component) whose saturation caps throughput; once it saturates, added load mostly turns into queueing and rising latency rather than more completed work.
Pf Apm Profiling Bottleneck CorrelationDifficulty 1
A load test shows client-observed p95 latency rising sharply after 500 concurrent users. What is the next reasonable step before concluding the server is the bottleneck?
- aImmediately report the server as the bottleneck since latency rose
- bStop the test, since rising p95 latency always means the test configuration (VU count, ramp-up) is wrong and must be redone
- cIncrease the number of virtual users further to double-check
- dCheck server-side metrics (CPU, GC pauses, DB pool wait, thread queue) for that same time window.✓
Explanation:Rising client-observed latency alone does not prove where the bottleneck is; it could be the server, the network, or even the load generator itself. Correlating with server-side metrics for that same time window is needed before attributing the cause.
Pf Apm Profiling Bottleneck CorrelationDifficulty 1
What does "saturation" mean for a resource such as a database connection pool during a load test?
- aThe resource has been restarted by the operating system, which resets its internal counters and connection state
- bThe resource is used at or near its maximum capacity, so new demand must wait.✓
- cThe resource has been permanently disabled by the load test
- dThe resource's response time has dropped to zero
Explanation:Saturation means a resource (CPU, connection pool, thread pool, etc.) is at or near 100% utilization; further requests must queue for it, which is what typically drives the latency increase observed by the load tool.
Pf Apm Profiling Bottleneck CorrelationDifficulty 2
Why can client-side latency alone be misleading when trying to find where a bottleneck is located?
- aIt only shows total round-trip time, without indicating which internal component contributed most of it.✓
- bClient-side latency numbers are always fabricated by load testing tools
- cClient-side latency is measured in a different unit than server-side latency, making comparison impossible
- dClient-side latency never changes regardless of server load
Explanation:Client-observed latency is an end-to-end measurement; it tells you something is slow but not which internal hop or component is responsible, which is why it needs to be correlated with internal (server-side) metrics to isolate the actual bottleneck.
Pf Apm Profiling Bottleneck CorrelationDifficulty 2
During a load test, the load generator's own host shows 95% CPU usage while the target server shows only 30% CPU usage, yet client-observed latency is high. What is the most likely explanation?
- aThe database is definitely the bottleneck
- bThe results should be trusted as-is since the server is clearly healthy and healthy server metrics always mean the client experience is accurate
- cThe network between load generator and server is definitely saturated
- dThe load generator itself may be the bottleneck, so the observed latency may not reflect the server's real behavior.✓
Explanation:If the load generator's host is CPU-saturated, it may fail to issue requests on schedule or accurately timestamp responses, producing misleading latency numbers that look like a server problem but actually originate from an overloaded client. This is a classic case of misattributing a client-side bottleneck to the server.