Get Free Quote

A 90-Day CloudOps Plan, Step by Step, for IT Coordinators

Healthcare IT

You have been told to get the organisation’s cloud house in order and you have ninety days. There is no clinical content here and no medical guidance: this is a plain, step-by-step plan for the administrative and technical side of running cloud-hosted systems, such as scheduling tools, document stores and staff portals. It follows the staged outline that the Ops Atlas from xOps School suggests for CloudOps, and it is written for IT coordinators who prefer instructions to theory.

Before you start, read the Atlas page that defines the CloudOps scope and the thirty, sixty and ninety day path. The page calls the discipline emerging and admits an editorial slant, so adjust the steps to your size and your obligations. Anything touching regulated data should be checked with your own compliance advisers.

Step 1 (days 1 to 3): decide what you are in charge of

Write one paragraph that says which cloud accounts and services are in scope and which are not. The Atlas itself draws boundaries: spending optimisation, security posture, policy authoring, code-defined infrastructure, networks, clusters, functions and host operating systems belong to neighbouring disciplines. Do the same for your organisation, and name an owner for each excluded area so nothing falls between desks.

Step 2 (days 3 to 10): build the inventory

Write down each cloud account with its owner, its reason for existing and a rough monthly cost. Compare your list with the provider’s own account listing and with the invoice. Any difference is a finding. Mark accounts you cannot explain and ask around; abandoned experiments are common, and each one is a ceiling shared, a bill paid and a door left unlocked.

For each important system, also draw up a two-column responsibility sheet: duties the provider carries on one side, duties left to you on the other. Keep it with your continuity papers.

Step 3 (days 10 to 15): fix where provider messages land

Create a shared, monitored address for provider notices, redirect the account contacts to it and name a duty owner for each week. Create a simple log: date received, summary, owner, due date and closing note. The Atlas lists notice closure among its measures, and it names routing signals as part of the first-month work for good reason. A missed retirement notice becomes an urgent problem at the worst moment.

Step 4 (days 15 to 30): read the estate against a baseline

Decide what every account should have: logging turned on, owner and environment labels, budget alerts and a named contact. Check each account against the list and record the gaps. Do not fix everything yet. The goal of this stage is evidence about how far apart your accounts are.

If you track household or team money with a routine, the approach in our piece on a monthly expense-tracking habit that lasts will feel familiar: define categories, record consistently and review on a fixed date.

Step 5 (days 30 to 45): set up a landing zone, or an interim template

A landing zone is a reusable recipe so that each fresh account starts from identical settings. Provider-supplied tooling can create accounts from it on request. If that is too large a step, write a checklist that a person follows for each new account and aim to automate it later. One caveat from the Atlas: restricting policies skip the top management account in AWS, which is meant to stay free of workloads.

Step 6 (days 45 to 55): turn on guardrails in report-only mode

Choose three: permitted locations, required labels and budget alerts. Run each in audit mode and gather what would have been blocked. Write down the blind spots before anyone asks. A location rule, for instance, can be unhelpful for things already running and for services that are global by nature, one platform’s location policy skips resource groups, and another’s covers creation but not later data processing. For labels, one platform may not list an unlabelled resource as a failure, and another does not pass labels from group to contents.

Step 7 (days 55 to 60): create the quota alarms

List the limits behind your most important services and set an alarm at a threshold such as 80%. Calculate the runway for each. The Atlas gives a reference sum: with a limit of 256, an 80% warning line, 184 in use and growth of 2 a day, there are 10.4 days until the alarm and 36 until the limit. Look for the traps it lists, namely reserved but idle capacity that still counts, special hardware families that start at zero and providers adjusting limits themselves.

Step 8 (days 60 to 70): design the exception path

Before enforcing anything, write the exception procedure: who to ask, what information is needed, who decides, how long the exception lasts and where it is recorded. Aim to decide within one working day. The Atlas lists guardrails with no exception route as a mistake, since people will otherwise route around the rule. For a human-scale parallel, see our checklist for preparing paperwork before a first bank visit, which shows how much smoother a process runs when every document and decision-maker is known in advance.

Step 9 (days 70 to 80): enforce, one group at a time

Switch the rules from reporting to enforcing, starting with one organisational unit or one low-risk account group, and watching the first week closely. Fix surprises before expanding. Bring the remaining groups in over the next days. Keep the exception log open and review it weekly.

Step 10 (days 80 to 90): rehearse a recovery and count the steps

Choose your most important system and simulate losing its primary location. Walk through the recovery plan and mark every step that requires the provider’s management layer: creating resources, changing policies or asking for capacity. The Atlas holds that the shorter the recovery target, the fewer such steps should remain, and that for the top tier none should. During a long regional incident it describes, requests for new machines failed for hours while those already running stayed healthy, which is the case for pre-built standby capacity. Record the result, assign fixes and set a date for the next rehearsal.

Step 11 (day 90): measure and report

Pull together the numbers: how many accounts were created from the template, how many carry the baseline, how many resources sit in approved locations, how much runway each key limit has and how many notices were closed. The Atlas offers a day-ninety marker of more than 80% baseline coverage, and longer-term targets of 95% or more for coverage measures and no quota-caused launch failures. Report your own results, including the gaps.

Common snags along the ninety days

Expect a few predictable delays. The inventory takes longer than planned because some accounts were opened by people who have since moved on. Owners reply slowly because the request looks like extra work, so explain that the register protects them from surprise bills and sudden shutdowns. The notice mailbox fills with low-value messages, so agree on a quick triage rule: anything mentioning retirement, deadline or action required gets an owner at once, and the rest is skimmed weekly. Audit reports look worse than expected on the first run; that is normal and exactly why the report-only stage exists. If progress stalls, shrink the scope, finish one account group completely and use it as the example for the others.

What to do with the result

There is a five-rung ladder in the Atlas, which it calls its own scheme: improvised, foundations, governed, measured and resilient. After ninety days you may reasonably say you are at foundations, working toward governed, where roughly nine accounts in ten are vended and 95% hold the baseline. Do not present it as a grade. Present it as a map: where you started, where you stand and which rung comes next.

Plan a quarterly review. Repeat the inventory, refresh the quota runways, close old exceptions and rehearse a different scenario. The first ninety days build the habits; the following quarters keep them alive.