EOXIS EOXIS
Home › Expertise › Performance & Observability
Expertise 01

Performance & Observability

Knowing, before go-live, whether the application will hold. And going all the way to the fix - not just to the report.

Discuss a campaign Related services
P95 response time against load, before and after tuning 500 users 5,000 users 6 s 3 s 0 Contractual threshold: 1 s Before: 5.8 s After: 0.8 s
Before tuning After tuning
What we bring

One measurement, one cause, one fix

  • Performance requirements written with the business, not inferred after the fact
  • Load scenarios faithful to real user journeys and real volumes
  • The test run correlated with application observability: you see the transaction, the query, the pool that saturates
  • A recommendation carried through to the code, then a re-run that proves it holds
  • A handover to your teams: next time, you can run it yourselves
Related services

Performance testing campaign

Scoping, scripting, ramp-up, endurance and robustness runs, report and action plan.

Application audit and tuning

Slow transactions, profiling, SQL, pools and caches, runtime - all the way to the fix.

Observability and APM

Platform selection, instrumentation, per-application dashboards, alerting that earns its keep, handover.

Usage scenario

From a go-live date to a verdict

1
You → EOXIS

"We are rebuilding the portal, go-live in ten weeks."

Half a day with the business and the operations team: critical journeys, expected volumes, contractual thresholds, test environment.

Deliverable: the test plan, costed
2
EOXIS

Scripts, data, load injectors, instrumentation

The journeys are scripted on your tooling or on ours - JMeter, OctoPerf, NeoLoad. The test environment is instrumented with APM so we can see inside while the run is going on.

Deliverable: the test platform, replayable
3
EOXIS

Ramp-up, endurance and robustness runs

Response times are read side by side with application metrics. The transaction that bends, the query that bends it, the pool that saturates: all on the same screen.

Deliverable: the test report and its causes
4
EOXIS → your developers

Fix, then re-run

The recommendation is carried through to the pull request. Once fixed, we run it again: it is the re-run that provides the proof, not the report.

Verdict: passed, passed with reservations, or failed

Tools we work with

Load injection
Apache JMeter · OctoPerf · NeoLoad · Tricentis · Gatling · k6

Observability and APM
IBM Instana · New Relic · Elastic APM · Datadog · Dynatrace · Grafana

With FlashOps

Our campaigns can be driven from FlashOps: the test run, the observability data and the offending code on one screen, and a report with a verdict drafted by the AI and reviewed by the engineer.

See the load testing connector →