MeasureUp Test Performance Analysis 3 — Questions and Answers
Question 1: Which queueing theory metric describes the average number of requests waiting in a queue at any given moment?
- Service time
- Utilization
- Mean queue length (Correct answer)
- Arrival rate
Correct answer: Mean queue length
Mean queue length (Lq) from Little's Law counts the average backlog of waiting requests.
Question 2: A database query plan shows a 'Sequential Scan' on a large table. What is the most likely performance fix?
- Increase connection pool size
- Add an index on the queried column (Correct answer)
- Partition the application servers
- Enable query result caching
Correct answer: Add an index on the queried column
Sequential scans on large tables are slow; adding an appropriate index allows the planner to use an index scan instead.
Question 3: What is 'coordinated omission' in load testing?
- Excluding outlier results from reports
- Load generators skipping requests when the system is slow, masking latency (Correct answer)
- Synchronizing clocks across test nodes
- Removing warm-up period data from results
Correct answer: Load generators skipping requests when the system is slow, masking latency
Coordinated omission occurs when a load tool waits for a response before scheduling the next request, hiding the true latency distribution under load.
Question 4: Which metric should you monitor to detect memory pressure before an out-of-memory event occurs?
- CPU steal time
- Page fault rate (Correct answer)
- Network packet loss
- Disk read throughput
Correct answer: Page fault rate
A rising page fault rate indicates the OS is swapping memory pages, signaling that RAM is insufficient for the workload.
Question 5: In APM (Application Performance Monitoring), what is a 'trace' composed of?
- A series of log lines from a single host
- A collection of spans representing operations in a single request (Correct answer)
- Aggregated metric histograms per service
- Network packet captures for one session
Correct answer: A collection of spans representing operations in a single request
A distributed trace is built from spans, each representing a timed unit of work across services handling one request.
Question 6: What does the Amdahl's Law formula predict about parallel performance scaling?
- Linear speedup with every added processor
- Speedup is bounded by the sequential fraction of the workload (Correct answer)
- Doubling threads doubles throughput always
- Network latency dominates at scale
Correct answer: Speedup is bounded by the sequential fraction of the workload
Amdahl's Law states that the maximum speedup is limited by the proportion of the program that cannot be parallelized.
Question 7: When performing a flame graph analysis, what does a wide flat top in the flame graph indicate?
- A function that calls many children quickly
- A function consuming a large share of CPU time (Correct answer)
- An I/O wait cycle
- A memory allocation hotspot
Correct answer: A function consuming a large share of CPU time
Wide horizontal bars in a flame graph represent functions where the CPU spends proportionally more time.
MeasureUp Test Performance Analysis 3
Which queueing theory metric describes the average number of requests waiting in a queue at any given moment?