Performance Benchmarking & Tuning Flashcards
7 cards from real CSI practice questions. Tap to flip, then mark Knew It or Still Learning — missed cards come back until you master them.
Read the first 7 Performance Benchmarking & Tuning flashcards as text
During performance tuning, what is the correct sequence for identifying and resolving bottlenecks?
Answer: Measure → profile → identify bottleneck → tune → re-measure
Effective tuning follows a data-driven loop: measure baseline, profile to locate the bottleneck, apply a targeted fix, then re-measure to confirm improvement.
A system integrator is asked to evaluate an application's scalability. Which test type directly measures how performance changes as the number of users grows?
Answer: Scalability test (ramp-up load test)
A scalability or ramp-up load test incrementally increases concurrent users while monitoring resource usage and response times to determine how well the system scales.
What is the primary risk of tuning a parameter without establishing a baseline benchmark first?
Answer: There is no way to determine whether the change improved or degraded performance
Without a baseline, there is no reference point to compare against, making it impossible to determine whether a tuning change had any positive, negative, or neutral effect.
Which caching strategy is most appropriate when the same expensive computation is requested repeatedly with the same inputs?
Answer: Memoization or result caching keyed on input parameters
Memoization stores the result of expensive function calls keyed by their inputs so repeated calls with identical parameters return cached results immediately.
In a distributed system benchmark, 'coordinated omission' refers to:
Answer: A measurement artifact where slow responses cause subsequent requests to be delayed, making latency appear artificially low
Coordinated omission occurs when a load generator waits for a response before sending the next request, inadvertently hiding queueing delays and under-reporting tail latency.
Which approach best reduces I/O latency when an application writes many small records to disk frequently?
Answer: Batching writes and using buffered I/O with periodic flushes
Batching small writes and using buffered I/O reduces the number of actual disk operations by coalescing multiple records into fewer, larger writes before flushing.
When reporting benchmark results to stakeholders, why is it important to include environmental context such as hardware specs and software versions?
Answer: Results are only reproducible and comparable when the test environment is fully documented
Benchmark results are meaningless without context; documenting hardware, OS, and software versions allows others to reproduce the test and makes cross-environment comparisons valid.