At a glance
Legacy sources: On-prem estates · Legacy Windows apps · Obsolete OS · Specialized peripherals · Regulated workloads
Modern targets: AWS · Azure · C# .NET · Golang · Rust · Cloud APIs
Typical timeline: 2 months to 18 months
Project range: $30,000 to $400,000
U.S. delivery: 100% U.S. citizen workforce, U.S.-soil delivery
What to know before you modernize
Why part of a system stays on-premises
Hybrid is justified by a specific constraint, and the constraint should be written down. Microsoft's description of Azure Local names the common ones: control systems and near real-time operations with extreme latency requirements, such as manufacturing execution systems; mission-critical operations that must keep running through a network outage, such as production lines; strict data sovereignty or regulatory rules that keep data under local control; and compute that simply has to remain on-premises. AWS adds unresolved physical dependencies, such as machines in a manufacturing plant, and data residency to its reasons to retain an application. Anything not on that list, such as a preference or a fear of change, is not a reason to keep running two environments.
Connecting the two sides: VPN, ExpressRoute, and Direct Connect
There are two kinds of link. A site-to-site VPN runs encrypted IPsec tunnels over the public internet. AWS Site-to-Site VPN gives each connection two tunnels for high availability, and Azure VPN Gateway recommends active-active mode with a tunnel to each gateway instance. A private circuit avoids the internet entirely: Azure ExpressRoute extends your network to Microsoft through a connectivity provider, and AWS Direct Connect links your router to an AWS location over fiber. Microsoft states that ExpressRoute offers more reliability, faster speeds, and consistent latencies than typical internet connections, and recommends two circuits in two peering locations for maximum resiliency. A VPN is quicker and cheaper to set up and suits modest, tolerant traffic. A private circuit takes longer to provision and suits steady or latency-sensitive traffic. Address ranges on the two sides must not overlap in either case.
One identity across the boundary
Users and services should not carry separate credentials for each side. The usual arrangement keeps Active Directory as the source of identity on-premises and synchronizes accounts to the cloud directory. For Azure, Microsoft Entra Connect and Microsoft Entra Cloud Sync do this; Microsoft recommends Cloud Sync for new setups, and notes that Azure AD Connect V1 was retired on August 31, 2022 and that Connect V2 needs Windows Server 2016 or newer. Plan for the failure case as well: a plant-floor station must still let its operator log in when the link to the cloud is down, so the on-premises side needs a directory it can reach locally, and cloud-issued tokens should be treated as a convenience for cloud services, not a dependency for local operation. Service accounts and API keys are inventoried and rotated as part of the same exercise.
Data synchronization and how it fails
Synchronization between sites is where hybrid systems break, so decide first which side owns each kind of record. Two writable copies of the same record create conflicts that no tool resolves correctly without a business rule. For the copy itself, change data capture reads the database transaction log, and replication services such as AWS Database Migration Service perform an initial load and then ongoing replication. The failure modes are predictable. A link outage builds a backlog that arrives as a burst. Messages are delivered more than once or out of order. A schema change on one side stops the feed. Clocks differ between sites. The remedies are plain: queue changes at the source and send them when the link returns, give every change a stable identifier so duplicates are ignored, include a version or sequence number, and monitor the age of the oldest unsent change.
Cloud services that run on your own premises
When a workload needs cloud tooling but cannot leave the building, both providers sell infrastructure installed on the customer's site. AWS Outposts is a fully managed service that extends AWS infrastructure, APIs, and tools to customer premises. AWS operates and monitors the capacity as part of an AWS Region, for workloads that need lower latency or local data processing. Azure Local is Microsoft's distributed infrastructure for customer-owned locations. It runs both modern and legacy applications, can operate connected or disconnected from the cloud, and uses Azure Arc as its control plane; it is priced per physical core on the machines. These products reduce the difference between the two environments, but they do not remove the need for a link, a hardware budget, and someone to own the site. For a single legacy application, a plain server and a VPN is often enough.
Operating and monitoring both halves
A hybrid system fails at the seams, so operations has to see both sides in one place. Azure Arc projects servers, virtual machines, and Kubernetes clusters hosted outside Azure into Azure Resource Manager, so they can be inventoried, governed with Azure Policy, and monitored with Azure Monitor through VM extensions. The Arc control plane itself carries no extra cost for servers, while Azure services used on them are billed at their normal rates. Where Arc is not used, the same result comes from running one telemetry standard on both sides: OpenTelemetry for .NET applications exports logs, metrics, and traces over OTLP to a single backend. Whatever the tooling, the alerts that matter in hybrid are link health, replication lag, and the age of the oldest queued message, and each needs a named owner on both sides of the boundary.
A phased path for a hybrid estate
Microsoft's migration planning guidance gives a workable order. Map dependencies first, and move directly dependent components together, since they need low latency between them. Document anything that cannot move, why it cannot, and what data it shares. Keep the time spent in a split state short, and where components can move later, record their connections and plan the end date. Start with simpler, lower-risk workloads and non-production environments to prove the process, then move critical systems after that success. For a legacy Windows application this usually means the hardware-facing component stays and gains a defined interface, such as a local service or message queue, so that everything behind it can be modernized and moved without touching the hardware. Each phase has its own rollback plan, tested before it is needed.
Modernizing with Hybrid Solutions
Hybrid is the right answer when something genuinely cannot move to the cloud: regulated data residency, hardware peripherals, deterministic latency requirements, or just the boring fact that a piece of equipment expects a serial port and an old PC under a desk. The cloud parts of the system get the elasticity, observability, and deployment discipline of cloud-native. The on-premises parts stay where they belong. Phased, not heroic.
Overcoming Legacy System Obstacles
Legacy applications often pose real integration challenges: 16-bit or 32-bit dependencies that resist 64-bit migration, peripheral hardware with no modern driver support, and operating systems past their support windows that nonetheless run mission-critical processes. Hybrid modernization isolates these constraints so they do not block the parts of the system that can move forward.
Advantages of Hybrid Solutions
Hybrid earns its complexity when the alternative is forcing a bad fit. A factory floor application talking to PLCs over a deterministic industrial protocol does not belong in a public cloud region halfway across the country. A reporting and analytics layer over the same data absolutely does. Hybrid lets each part live where it works best. The real engineering work is in the seams: the integration patterns, the data synchronization, and the operational handoffs between environments.
Comprehensive Hybrid Solutions for Modernization
The starting question is what actually has to stay on-premises and why. From there: modern application code (C# .NET, Go, or Rust), cloud backend APIs on AWS or Azure where they make sense, cross-platform clients where users are mobile, and event-driven integration between the two halves. For programs requiring U.S. citizen workforce and U.S.-soil delivery, we operate entirely under those constraints. We are not currently DDTC-registered for ITAR, and we are honest about that during scoping.
Start with an assessment for a hybrid estate
For a hybrid candidate the assessment inventories what decides the boundary: every component with its operating system and runtime version, the hardware it talks to and over which interface (serial, USB, vendor SDK, industrial protocol), 32-bit and 16-bit dependencies, the latency each interaction can tolerate, and which data is regulated or subject to residency rules. It also covers the database and any existing synchronization or file transfers, authentication and service accounts, licensing terms that tie software to a site or machine, network paths and bandwidth, and the current monitoring and recovery procedures. Everything is read from source, configuration, and documentation; nothing is run and nothing in production is touched. The result is a written recommendation of what stays, what moves, and how the two halves connect. The fee is quoted after the free 1-hour consultation and creditable toward a subsequent modernization project.
Deliverables
- Modernized application in C# .NET, Go, or Rust
- Cloud backend APIs on AWS or Azure
- Cross-platform access (desktop, tablet, iOS, Android)
- Re-architected around business domains (microservices, CQRS)
- 100% U.S. citizen, U.S.-soil delivery
Common questions
What is a hybrid cloud architecture?
A hybrid architecture runs part of a system on-premises and part in a public cloud, connected by a network link and shared identity and data flows. It is used when some components cannot move, for example because of latency, regulation, or attached hardware, while others benefit from cloud capabilities such as elastic capacity and managed services.
When does it make sense to keep part of a system on-premises?
When a control system needs very low and predictable latency, when operations must continue through a network outage, when regulations or contracts require data to stay under local control, or when software depends on attached hardware. Those constraints should be documented per component. A system without such a constraint is usually simpler to run entirely in one place.
Should we use a VPN or a private circuit like ExpressRoute or Direct Connect?
A site-to-site VPN uses encrypted tunnels over the internet. It is quick to set up and fits modest traffic. ExpressRoute and AWS Direct Connect are private connections that avoid the public internet, take longer to provision, and cost more, and they fit steady or latency-sensitive traffic. Many systems start with a VPN and add a private circuit once traffic justifies it.
What happens to the on-premises system when the internet connection goes down?
It should keep working. The local side needs its own identity source, its own copy of the data it requires, and a queue that holds outgoing changes until the link returns. Design for this explicitly and test it by disconnecting the link. Systems that stop whenever the cloud is unreachable are not truly hybrid, only remote.
How do you keep on-premises and cloud databases in sync?
Assign one side as the owner of each kind of record, then copy changes with change data capture or a replication service. Changes are queued at the source, carry stable identifiers so duplicates are ignored, and are applied in order. Monitor replication lag and the age of the oldest unsent change, because a quiet failure there is the most common problem.
Do we need Azure Arc to run a hybrid system?
No. Azure Arc is one option for managing servers and Kubernetes clusters outside Azure from the Azure portal, with policy, monitoring, and updates. A hybrid system also works with standard tooling and one shared telemetry format on both sides. Arc is most useful when you already operate on Azure and have many on-premises machines to govern.
How long does a hybrid modernization take, and what does it cost?
The bands on this page apply: 2 months to 18 months, and $30,000 to $400,000. A single application with one hardware boundary sits at the low end, and a multi-site estate with synchronization and several integrations at the high end. The price and schedule are fixed after the Legacy Software Assessment, whose fee is quoted after the free 1-hour consultation and creditable toward a subsequent modernization project.
Sources
- What is Azure Local: Use cases (latency, continuity, sovereignty), connected/disconnected, Arc control plane, per-core pricing
- AWS migration strategies (7 Rs): Reasons to retain: residency, physical dependencies
- Azure ExpressRoute overview: Private connection, not over public internet, resiliency recommendation of two circuits in two locations
- AWS Direct Connect: Dedicated fiber connection bypassing internet providers
- AWS Site-to-Site VPN: IPsec, two tunnels per connection for high availability, non-overlapping CIDR recommendation
- Azure VPN Gateway site-to-site tutorial: Site-to-site IPsec; active-active recommendation; non-overlapping ranges
- Microsoft Entra Connect V2: Azure AD Connect V1 retired Aug 31 2022; V2 requires Windows Server 2016+; Cloud Sync recommended
- AWS Database Migration Service: Initial load plus ongoing replication
- Strangler Fig pattern (Azure Architecture Center): Change data capture reads the transaction log
- What is AWS Outposts: AWS-managed infrastructure on customer premises for low latency and local data processing
- Azure Arc overview: Projects servers/Kubernetes into Azure Resource Manager; extensions; no extra cost for control plane features
- .NET observability with OpenTelemetry: Logs, metrics, traces over OTLP
- Plan your migration (Cloud Adoption Framework): Dependency mapping, split-environment guidance, sequencing, rollback
Not sure where to start?
Start with a fixed-scope Legacy Software Assessment: a read-only review of your application’s source, database, configuration, logs, and architecture, ending in findings with evidence and a written recommendation. Quoted after the free 1-hour consultation; creditable toward a subsequent modernization project.