A practical 2026 checklist for CVAD and Citrix DaaS—built around what users see under concurrent load.
Direct answer: To load test Citrix Virtual Apps and Desktops (and Citrix DaaS), script the real client path, measure on the glass, and run a logon storm separately from a density / sustain test. CitraTest VU® drives StoreFront / Workspace → HDX session → published app with image recognition and no agents on production servers.
Step-by-step
- Freeze the image and path — note CVAD/DaaS version, Gateway vs lab shortcut, StoreFront/Workspace URL, and the published desktop or app under test.
- Script the real workday GUI — launch client, authenticate (include MFA path if production uses it), open the resource, complete a short business action users care about.
- Baseline one user on the glass — time login-to-shell / first paint and the app action before any concurrency.
- Run the logon storm — steep concurrent ramp; record p95/p99 logon, fail rate, profile IOPS. Stop. Do not blend into density yet. (Why the split matters)
- Run density / sustain — hold target concurrent users; watch in-session UX and session-host headroom.
- Stop at a defensible line — concurrent users both tests support with tails a CAB can accept.
- Re-test after change — image, GPO, FSLogix/profile, Broker, or autoscale plan changes need a fresh storm + density pair.
What to measure
- Login-to-usable desktop / first paint of published app
- p95/p99 tails and fail rate (not only averages)
- Host CPU/memory and profile storage under load
- Whether Director is green while the glass is slow
Tools and deeper reading
FAQ
Is protocol-only Citrix scripting enough?
Useful for infra signal. It can miss slow on-screen apps. Pair or prefer on-the-glass for UX capacity.
Do I install agents on VDAs?
Not with CitraTest VU. Generators sit outside production.
Same method for Citrix DaaS?
Yes — same storm/density split and glass metrics; autoscale changes how you interpret density.