Cloud & Enterprise IT
Cloud Architecture & Migration
Planning and executing moves to public cloud with tested cutovers and no surprises.
What this covers
Cloud migration for a public agency is a sequence of decisions: which workloads move, in what order, to which service model, and under what controls. HQT helps agencies make those decisions with evidence and then executes the plan.
We work with Amazon Web Services, Microsoft Azure, and Google Cloud, including their government regions where an agency's data classification requires it.
Typical activities
- Application and infrastructure discovery, including dependencies that are undocumented
- Disposition of each workload: rehost, replatform, refactor, replace, retire, or retain
- Target architecture, including account or subscription structure, networking, and identity
- Wave planning with cutover windows, rollback criteria, and communications
- Data migration with validation of row counts, checksums, and application behavior
- Post-migration tuning of performance and cost
What you receive
A migration plan with an inventory, disposition decisions, and a wave schedule; infrastructure as code for the target environment; cutover runbooks; and a post-migration report with open items and recommendations.
Considerations for public agencies
Data residency, records retention, CJIS and HIPAA controls, and procurement rules for cloud consumption all shape the plan. We document the shared responsibility split between the agency, the cloud provider, and HQT so that audit questions have clear answers.
Conditions that put migration on the agenda
The decision is usually forced by a date or a constraint rather than chosen in the abstract. Conditions that commonly bring it forward:
- Server hardware, storage, or hypervisor licensing is approaching end of support, or a renewal quote arrives that no longer resembles prior years.
- A data center lease, colocation agreement, or building disposition has a date attached to it.
- The disaster recovery site exists on paper but has never carried production users.
- A line of business vendor announces that future releases will be hosted only, and the on-premises version moves to extended support.
- The staff who understand the storage array, the backup jobs, or the network core are approaching retirement, and the positions are hard to fill.
Where several of these are true at once, sequencing matters more than platform choice. The date that cannot move sets the first wave, and the rest of the schedule is built backward from it.
Failure modes to design against
- Everything is rehosted unchanged, so the agency runs the same oversized servers at a monthly rate instead of on depreciated hardware.
- The target environment is assembled by hand in a web console and cannot be rebuilt the same way twice.
- Nothing is tagged, so finance receives a single bill that cannot be allocated to departments or funds.
- Rollback criteria are never written down, so the decision to continue during a troubled cutover is made under time pressure by whoever is in the room.
- The source environment stays powered on indefinitely after go live, and the agency pays for both.
- Non-production environments run around the clock because no one owns the schedule.
What to require in a solicitation
Requirements that can be tested at acceptance rather than answered in narrative:
- Discovery priced as its own deliverable, with a steady state cost estimate attached to each workload before migration waves are priced.
- Agency ownership from the first day. The cloud accounts or subscriptions, the billing relationship, DNS, and the repository holding infrastructure code belong to the agency during the project, not at closeout.
- Cutover rehearsal at realistic data volume, plus a named agency official with authority to stop a cutover and call the rollback.
- Knowledge transfer with acceptance criteria, including a period in which agency staff perform patching, changes, and restores while the contractor observes.
- Exit terms: export formats, timeline, and the cost to leave.
- Migration labor and any ongoing managed service priced separately, so the operations decision stays independent of the move.
- The authorization basis stated by the agency rather than left to the bidder. Requirements differ by data type and jurisdiction: FedRAMP is a federal program that state and local solicitations sometimes reference, some states run their own cloud authorization programs, and criminal justice and health data carry their own agreements and personnel requirements. Confirm what applies before release.
Measuring the result after cutover
Take the baseline before the first wave moves, because it cannot be reconstructed afterward. Measures worth reporting:
- Monthly run rate against the forecast that justified the move, broken out by the tags finance needs, with variances explained rather than absorbed.
- Recovery time and recovery point demonstrated by an actual restore rather than by a design document.
- Decommissioning completed: hardware powered down, licenses and support contracts ended, media sanitization recorded.
- The share of the environment that can be rebuilt from code with no manual steps.
- Whether agency staff have completed a patch, a change, and a restore without contractor assistance.
Effort per wave is worth tracking as well. If later waves consume about what the first one did, discovery and design work is being repeated rather than reused.
Working with HQT on this
HQT scopes this kind of work as a fixed-price or time-and-materials engagement with a written statement of work, named staff, and a defined acceptance process. We are candid about what we can take on and prefer a well-run project to an oversized one. Contact us to discuss a requirement, a draft scope, or an upcoming solicitation.
Frequently asked questions
How long does a cloud migration take for a city or county?
The schedule is set by the number of workloads, the dependencies between them, and how many change windows the agency can absorb without interrupting service. Discovery tends to determine the timeline more than the technical work does, because undocumented integrations surface there. Waves are commonly scheduled around billing cycles, tax deadlines, elections, and fiscal year end, so an agency with a busy calendar will spread the work across more than one budget period.
Will moving to the cloud reduce our IT costs?
Not on its own. Savings come from retiring systems nobody uses, rightsizing what remains, replacing self-managed servers with managed services, and shutting down non-production environments outside business hours. Rehosting servers at their existing sizing with no cost controls can cost more per month than continuing to run hardware the agency has already paid for. The change also shifts spending from capital to operating, which affects how the agency budgets and appropriates.
Can we move CJIS or HIPAA regulated data to a public cloud?
Generally yes, provided the service and region are appropriate for the data and the required agreements are in place. For protected health information that means a business associate agreement where the agency acts as a covered entity or business associate; for criminal justice information the CJIS Security Policy sets expectations for providers and their personnel, including the security addendum and screening. The provider covers part of the control set and the agency retains the rest, including access control, encryption decisions, auditing, and personnel screening. Confirm the specifics with your state CJIS Systems Agency, your privacy officer, and counsel before committing to a design, because policy versions and state supplements change.
Do we have to move everything, or can some systems stay on premises?
Retain is a legitimate disposition, and a hybrid end state is common. Systems tied to physical equipment such as SCADA and building controls, licensing that does not permit hosting, and appliances that must sit on a local network often stay where they are. Weigh the tradeoff before deciding: if a handful of systems keep a data center or communications room open, the agency continues to pay for the facility, cooling, and support around them.
What happens if a cutover fails?
The plan should define a point of no return, the criteria for calling a rollback before that point, and the person with authority to call it. Up to that point the source environment stays available and unaltered, which is why decommissioning is scheduled well after go live rather than during the window. After the point of no return, recovery means reconciling data written since the switch, so the rehearsal should establish how long a reversal would take and what would be lost.