Traffic and dependency map
Map sign-in, answer storage and receipt delivery. Use class timings and recent measurements to define a representative load test.
Timed online assessments can bring a full cohort into a portal at once. GreyBath helps teams rehearse that traffic, verify each submission and plan releases around the assessment window.
See the assessment rehearsal
01 / What we build
Map sign-in, answer storage and receipt delivery. Use class timings and recent measurements to define a representative load test.
Run a non-production rehearsal with synthetic attempts. Compare submitted, acknowledged and pending counts and trace one retry through to a single receipt.
Agree the release window, the conditions for pausing it and the evidence needed to proceed. Give support a view of pending attempts and a clear escalation owner.
City context, honestly stated
Delhi’s Directorate of Training and Technical Education lists certificate and short-term vocational courses across government and private ITIs. The example below concerns a fictional independent training provider that runs its own online assessments. GreyBath delivers the engineering remotely.
Three fictional classes of 40 learners each have the same 11 a.m. deadline. Use synthetic accounts to reproduce the sign-in burst and the final submission burst, then record how long the app takes to acknowledge each attempt. Agree the acceptable wait with the training team before reading the test results.
In a rehearsal, the portal saves one attempt but the client loses its connection before the receipt appears. Let the learner retry using the same attempt reference. The trainer should see one recorded attempt and the learner should receive the same receipt after reconnecting.
A pending release is ready while the queue still contains unconfirmed submissions. Keep the current version serving learners, pause the deployment and give support a list of attempts awaiting confirmation. Resume only after the named owner checks the backlog and the training team clears the next window.
Where it fits
Bring the people who set assessment times, maintain the portal and respond when someone cannot confirm a submission.
Share cohort sizes and deadline rules. Agree the message to show learners if a receipt is delayed.
Review queue and datastore capacity, staging parity, and the order in which the app can be safely updated.
Use an attempt reference to tell whether an answer was recorded, is still pending or needs an authorised review.
Schedule deployments away from assessment deadlines and keep a release paused while accepted attempts await reconciliation.
How we work
Follow a sample account from sign-in to submission receipt and identify every dependency that could slow or interrupt it.
Use an approved staging environment and synthetic answers. Compare the recorded count with the client receipts and investigate every mismatch.
Simulate a backlog before a planned release. Confirm the deployed version stays stable and that support knows what to check before the next attempt.
Start with one portal, one timed workflow and one representative peak. Agree the test environment and data before setting load or recovery targets.
Discuss your projectUse the expected number of active learners and the minutes when they sign in or submit. Confirm planned test load with the application owner.
Identify which system records the official attempt, when it is safe to retry and which signal tells the learner the attempt was accepted.
Name who can pause a deployment, confirm pending attempts and decide whether to roll back or fix forward.
Yes. Start with application measurements and the current provider’s available controls. Consider migration only if the evidence and constraints support it.
Yes. Use synthetic accounts and answers in an approved environment. Match the representative traffic pattern without copying private assessment material.
Agree wording for the pending state and a reference support can look up. If the first attempt was accepted, retry can show the same receipt.
Use actual cohort deadlines and observed request patterns to agree a workload, then compare the rehearsal with the service targets your team sets.
Bring a recent timetable, expected concurrent cohort sizes and a mock submission receipt. Include any known times when support sees delayed confirmations.
Bring a recent assessment schedule