Sample questions
Pf Distributed Load GenerationDifficulty 1
Why can a single machine typically not generate the load needed to simulate tens of thousands of concurrent virtual users?
- aLoad testing tools refuse to run more than a few hundred virtual users on any single process by design
- bThe target system always rejects traffic coming from a single source IP address
- cA single machine has finite CPU, memory, sockets and bandwidth, capping realistic virtual user counts✓
- dOperating systems never allow more than one TCP connection to be opened from a single machine at a time, no matter how it is configured or tuned
Explanation:A single load generator is bounded by its own finite resources (CPU cycles to run scripting logic, memory per VU, available ephemeral ports/sockets, and NIC bandwidth), which is why very high virtual-user counts require distributing the generation across multiple machines.
Pf Distributed Load GenerationDifficulty 1
In a distributed load generation architecture, what is the general role of the coordinator (controller) node?
- aIt orchestrates workers: starts/stops them, distributes the test plan, aggregates results✓
- bIt is the only node that actually sends requests to the system under test, while every worker simply idles and does nothing throughout the run
- cIt replaces the need for any monitoring of the system under test during the run
- dIt automatically rewrites the test script to remove any bugs found by the workers
Explanation:The coordinator distributes the workload definition to worker nodes, controls the test lifecycle (start/stop/ramp), and aggregates the metrics each worker reports back — the actual traffic generation happens on the workers.
Pf Distributed Load GenerationDifficulty 2
In Locust's master-worker mode, which command correctly starts the master process?
- alocust -f locustfile.py --node-type=master
- blocust -f locustfile.py --coordinator
- clocust -f locustfile.py --controller
- dlocust -f locustfile.py --master✓
Explanation:Locust uses the --master flag to start a process in master mode, which then waits for worker processes (started with --worker) to connect before distributing load generation across them.
Pf Distributed Load GenerationDifficulty 2
When starting a Locust worker process that must connect to a master running on host 10.0.0.5, which flag correctly points the worker to that master?
- alocust -f locustfile.py --worker --target-host=10.0.0.5
- blocust -f locustfile.py --worker --master-host=10.0.0.5✓
- clocust -f locustfile.py --worker --connect=10.0.0.5
- dlocust -f locustfile.py --worker --upstream=10.0.0.5
Explanation:Locust workers use --master-host (with the default port, or --master-port if it differs) to specify where the master process is running so they can register and receive their share of the workload.
Pf Distributed Load GenerationDifficulty 2
Why must the load generator machine's own CPU and network usage be monitored during a distributed load test, not just the system under test?
- aA saturated generator cannot produce the intended load, silently invalidating results✓
- bMonitoring the load generator machine is required by law in most countries that regulate performance testing activities
- cThe generator's CPU usage directly becomes part of the system under test's reported response time
- dGenerators never consume meaningful CPU or bandwidth, so this monitoring is only a formality
Explanation:If the load generator itself becomes CPU- or network-bound, it cannot issue requests at the intended rate; the test would then be measuring the generator's limits rather than the target system's behavior, giving misleadingly optimistic results.
Pf Distributed Load GenerationDifficulty 2
Why is it generally recommended NOT to run the load generator on the exact same host as the system under test?
- aBecause load testing tools refuse to start if the target hostname resolves to localhost
- bBecause the generator and the system under test would compete for the same CPU, memory and network resources, contaminating the results✓
- cBecause it would make the test run faster than any distributed setup, which is considered unrealistic
- dBecause HTTP requests cannot be sent to
localhost or 127.0.0.1 by most load testing tools
Explanation:Colocating the generator and the target means they share and compete for the same hardware resources, so a slowdown could come from the generator starving the target (or vice versa) rather than reflecting real-world behavior — this contaminates the measurement.