5 exercises on the language SREs, DevOps engineers, and backend developers use when discussing performance issues, observability, and incident management.
Core performance vocabulary patterns
detect → investigate → fix a memory leak
identify / resolve a bottleneck
collect → monitor → alert on metrics
reduce / improve / measure response time
flag → triage → resolve a performance incident
0 / 5 completed
1 / 5
An SRE writes in a post-mortem: "During last night's incident we ___ a memory leak in the payment service." All three options are grammatically correct — which verb best fits the discovery stage?
Detect — the standard verb for discovering a performance problem:
In SRE and performance engineering, there is a natural progression of verbs that describe the lifecycle of an issue:
detect / spot / notice → discovery phase
"Monitoring detected a memory spike." — system found it
"The team spotted a memory leak in the logs." — human found it
investigate / diagnose / analyse → root cause phase
"We investigated the memory leak using heap dumps."
"The profiler helped us diagnose the leak."
reproduce → confirmation phase
"We reproduced the leak in a staging environment."
"We fixed the memory leak by closing unclosed streams."
"The patch addressed the leak in the connection pool."
introduce / cause / trigger → origin verbs
"A recent refactor introduced a memory leak."
"The unbounded cache caused the leak."
Key distinction: "Detect" implies the problem was found; "introduce" means it was caused; "resolve" means it was fixed. Using the wrong verb in a post-mortem can completely change the meaning of the timeline.
2 / 5
A backend engineer is profiling an API and says: "I found a ___ in the database query layer — it's causing 3-second delays." Which word completes this sentence most naturally?
Bottleneck — the essential performance vocabulary word:
A bottleneck is a point in a system where the flow of work is constrained, causing the whole pipeline to slow down — like the narrow neck of a bottle limiting how fast liquid pours out.
Core collocations with "bottleneck":
identify / find / spot a bottleneck — discover where the constraint is
analyse / profile to find a bottleneck — use tools to locate it
resolve / eliminate / remove / address a bottleneck — fix the constraint
create / introduce / cause a bottleneck — generate the constraint
"The database is / becomes a bottleneck" — describe the cause
Compare similar terms:
bottleneck — a throughput constraint at a specific point (most general)
deadlock — two threads/processes waiting for each other indefinitely (specific concurrency bug)
memory leak — memory is allocated but never freed, causing gradual resource exhaustion
hotspot — a code path that consumes disproportionate CPU/memory (profiling term)
contention — multiple threads competing for the same resource (locks, connection pool)
In context: "The N+1 query pattern creates a bottleneck at scale." / "We identified the bottleneck using APM traces."
3 / 5
In a DevOps team meeting, a team member says: "We need to ___ latency and error rate metrics across all services." Which verb describes the ongoing process of watching metrics in real time?
Monitor — the verb for the continuous observability practice:
The metrics lifecycle involves several distinct verbs, each describing a different stage:
1. Instrument / instrument for / add instrumentation — add code to generate metrics
"We instrumented the service with Prometheus client library."
2. Collect / scrape / gather — fetch raw metric data
"Prometheus scrapes / collects metrics every 15 seconds."
"The agent gathers system metrics every minute."
3. Store / persist / ingest — write to a time-series database
"Grafana visualizes the metrics." / "We graphed the p99 latency."
5. Monitor / observe / track / watch — the ongoing human/automated practice
"We monitor error rates with Datadog."
"The on-call engineer tracks the SLOs in real time."
6. Alert on / alert when — trigger notifications
"We alert on p95 latency above 500ms."
"PagerDuty fires when the error rate exceeds 1%."
7. Analyse / diagnose — interpret metrics to find root cause
"We analysed the metrics and found the issue in the DB tier."
4 / 5
A performance engineer presents quarterly goals: "Our target for Q3 is to ___ the API response time from 450ms to under 200ms." Which verb is most natural here?
Reduce — the standard engineering verb for bringing metrics down:
When talking about performance improvements, the choice of verb matters for precision:
Reduce /rɪˈdjuːs/ — most common verb in engineering context
Implies deliberate effort to make a number smaller
measure / benchmark / profile — observation verbs, not direction-specific
5 / 5
An incident manager sends an update: "A performance degradation was ___ by an automated alert at 03:12 UTC." Which verb correctly describes how an alert system notified the team?
Flag — the standard verb for when monitoring tools surface an issue:
Incident management has a rich vocabulary of verbs with precise meanings:
Detect / flag / surface / raise — how the issue is first noticed:
"The alert flagged an anomaly in response time." — monitoring system surfaced it
"Datadog detected a spike in p99 latency." — system-level discovery
"The issue was surfaced by a customer report." — user-reported
"An automated check raised a degradation event."
Trigger — used for what causes the alert to fire, not the incident itself:
"High CPU usage triggered the alert." ✅ — CPU triggered the alert "The incident was triggered by the alert." ❌ — backwards; the incident triggered the alert
Escalate — moving an incident to higher-severity or more senior responders:
"After 15 minutes without resolution, the incident was escalated to senior SRE."