Why performance testing is skipped and why it shouldn’t be
Performance testing is often cut to save time before go-live. The result: a VBCS application that works perfectly with 1 concurrent user collapses under 50 concurrent users, or an OIC integration that processes 100 records fine takes 8 hours on the first production run with 50,000 records.
Performance issues found in production are 10x more expensive to fix than those found in testing.
What to performance test
OIC scheduled integrations: test with production-scale data volumes. If the integration processes daily transactions and Monday’s load is 10x a normal day, test with Monday-scale volumes. Key metric: total runtime — it must complete within its scheduled window.
OIC real-time integrations: measure response time under concurrent load. Key metric: 95th percentile response time under expected peak concurrency.
VBCS applications: test page load time with full data sets (not test environments with 100 rows), and concurrent user load. Key metrics: page load <3 seconds, form submission <2 seconds.
BIP reports: test with production-scale data. Run the report without caching first to get the worst-case time. Key metric: report generation <30 seconds for interactive; longer acceptable for scheduled.
Tools for load testing
Oracle Performance Testing Cloud Service (OPTS): Oracle’s managed load testing platform, integrates with OCI.
JMeter: open-source, widely used. Configure HTTP requests to simulate OIC REST trigger calls or VBCS API calls. Run scenarios with 50-500 concurrent virtual users.
Postman’s collection runner: for API endpoint performance — not full load testing but useful for single-endpoint response time benchmarks.
Baseline before optimising
Run baseline tests, document results, optimise, retest. Don’t skip the baseline — you need it to prove the optimisation worked and to establish what “acceptable” means.
SLA definition
Before performance testing, define what “good” looks like:
- VBCS page load: <3 seconds for 95% of users under peak load
- OIC REST integration: <2 seconds for 95th percentile
- Scheduled batch: complete within 4-hour window with 20% buffer
Test against these criteria — not against a vague sense of “feels fast.”