> For the complete documentation index, see [llms.txt](https://docs.vega365.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.vega365.com/redundancy-management-framework/solution-blueprints/redundancy-across-subnets.md).

# Redundancy Across Subnets

DR designs need not be constrained to a single subnet.

### Overview

<div align="center"><figure><img src="https://73860689-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FK95oxUATgzuZZ2VhRoDw%2Fuploads%2Forr4IBrjXfJxuiqyafVk%2Fimage.png?alt=media&amp;token=22092c6b-3eea-449b-8f5d-7ddc137d451f" alt="" width="563"><figcaption></figcaption></figure></div>

Milestone XProtect's native Management Server Failover is constrained to a single subnet. This constraint prevents customers from designing disaster recovery (DR) architectures in which the primary and secondary Management Servers reside in geographically separate data centers on different network segments.

The Vega Redundancy Management Framework (RMF) removes this constraint. RMF enables resilient Management Server architectures across subnets, without Windows Server Failover Clustering, SQL clustering, or shared database infrastructure between sites. The same cross-subnet design also carries recorded video between sites through **Media Backfill**, restoring gap-free archives on both sides after any temporary outage.

### The Subnet Boundary Problem

#### Native Management Server Failover requires a single subnet

Milestone's native Management Server Failover is built on a virtual IP address shared by the primary and secondary Management Server nodes. Because the virtual IP relies on Layer 2 adjacency and low-latency database replication, both nodes must belong to the **same subnet**. This is a documented prerequisite in the Milestone XProtect administrator manual.

The implication for DR design is direct: if the secondary Management Server is hosted in a different data center on a different subnet — a normal requirement for geographic redundancy — native Management Server Failover **cannot be deployed**.

#### Windows Server Failover Clustering is not a free alternative

Customers who attempt to work around the subnet boundary by adopting Windows Server Failover Clustering (WSFC) for the Management Server face a choice between two clustered SQL paths, each with its own cost.

**Path 1: SQL Server Failover Cluster Instance (FCI).** FCI provides instance-level high availability and works on SQL Server Standard Edition for a two-node cluster. However, FCI requires **shared storage** — a SAN, iSCSI target, or Storage Spaces Direct cluster — that is accessible to both nodes. This shared storage is itself a single point of failure, is a significant capital expense, and must be designed with its own redundancy. The Management Server tier appears redundant, but the storage tier underneath it is not.

**Path 2: SQL Server Always On Availability Groups (AGs).** AGs remove the shared-storage requirement by maintaining synchronized replicas on each node's local storage. However, **SQL Server Standard Edition supports only Basic Availability Groups**, which are limited to one database per AG, two replicas, and no readable secondary. Because XProtect's Management Server uses multiple SQL databases (Surveillance, Log Server, Identity Provider, Incidents, and others) that must fail over coherently as a unit, a production AG design for XProtect effectively requires **SQL Server Enterprise Edition** with advanced Availability Groups — a meaningful increase in licensing cost.

<figure><img src="https://73860689-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FK95oxUATgzuZZ2VhRoDw%2Fuploads%2Fsi4YfegUXVym3mlM0FpX%2Fimage.png?alt=media&amp;token=3d8b03c1-0425-4915-a360-69476ae02424" alt="" width="563"><figcaption></figcaption></figure>

Either path also requires:

* A domain-joined Windows Server Failover Cluster, quorum witness, and cluster networking.
* Identical OS and SQL Server patch levels across nodes, with cluster-aware patching procedures.
* Windows Server Datacenter Edition if the design exceeds two cluster nodes, since Windows Server Standard Edition supports only a two-node failover cluster.

And neither path removes the underlying architectural objection:

* **The database remains a shared failure domain.** Whether through shared storage (FCI) or coordinated replicas (AG), both cluster nodes operate on the same logical database. Failure, corruption, or compromise of that shared database affects both nodes simultaneously. Redundancy at the application tier is undermined by coupling at the data tier.
* **The subnet boundary is not cleanly eliminated.** Multi-subnet failover clustering is supported, but it adds DNS, IP, and client-driver dependencies rather than removing the subnet as a design dimension.

The alternative to the subnet constraint, in short, is a forced trade-off — shared storage or Enterprise licensing — layered atop substantial WSFC operational complexity, while preserving the shared-database failure mode that good DR design seeks to eliminate.

### The RMF Approach

#### Independent Management Servers, consistent configuration

RMF is built on a different design principle:

> Management Servers do not need to share a database. They need consistent configuration.

Each site in an RMF deployment runs its own **independent** Management Server, with its own SQL database, on its own subnet. There is no shared virtual IP, no clustered SQL instance, and no replicated database between sites. The two sites are fully isolated at the data tier.

What RMF provides is **configuration consistency** between the independent sites. The RMF SmartSync Service monitors configuration objects — cameras, hardware, users, roles, rules, views, and related entities — on the primary site and applies corresponding changes to the secondary site through the Milestone MIP SDK APIs.

#### API-based synchronization, not database replication

Because synchronization is performed through Milestone's published APIs rather than through raw database replication, the RMF approach has several properties that are important for cross-subnet DR:

* **No subnet dependency.** API calls operate at the application layer over standard TCP. Any network path that permits authenticated API communication between sites is sufficient. There is no Layer 2 adjacency requirement.
* **No SQL clustering.** Each site runs its own SQL Server instance, typically Standard or Express Edition as appropriate for the deployment size. There is no clustered SQL backend to license, deploy, or maintain.
* **No shared database failure mode.** A failure, corruption, or compromise of the database at one site does not propagate to the other.
* **Validated synchronization.** RMF reads configuration objects via the API and writes the corresponding objects to the secondary site. Malformed or invalid configuration fails validation rather than being copied byte-for-byte into the secondary database.

#### Each site operates independently

Each site in an RMF deployment is a complete, standalone XProtect installation. It can be operated, audited, upgraded, and recovered independently of its peer. This isolation is the foundation for both the cross-subnet flexibility and the architecture's cybersecurity posture: there is no shared attack surface between sites, and lateral movement between sites via database or cluster channels is not possible.

#### Media Backfill Across the Subnet Boundary

<figure><img src="https://73860689-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FK95oxUATgzuZZ2VhRoDw%2Fuploads%2F37d8ZWrRwgsESYIwwR7k%2Fimage.png?alt=media&amp;token=9447ee39-5d1e-4279-9417-f49f3bb22266" alt="" width="563"><figcaption></figcaption></figure>

Cross-subnet redundancy is not only a configuration-plane capability. **RMF Media Backfill carries recorded video across the subnet boundary** to repair gaps in either site's media archive — the same architectural property that lets RMF synchronize configuration across subnets also lets it synchronize media across them.

### Architectural Comparison

| Capability                                   | Native Management Server Failover | WSFC + SQL FCI                                        | WSFC + Always On AG                                                                                                          | **Vega RMF**                                          |
| -------------------------------------------- | --------------------------------- | ----------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------- |
| Cross-subnet Management Server redundancy    | No                                | Possible (multi-subnet WSFC)                          | Possible (multi-subnet WSFC)                                                                                                 | **Yes**                                               |
| Cross-subnet media backfill                  | No                                | No                                                    | No                                                                                                                           | **Yes**                                               |
| Requires same subnet                         | Yes                               | Not strictly, but adds DNS / IP / driver dependencies | Not strictly, but adds DNS / IP / driver dependencies                                                                        | **No**                                                |
| Requires Windows Server Failover Clustering  | No                                | Yes                                                   | Yes                                                                                                                          | **No**                                                |
| Requires shared storage                      | No                                | **Yes (SAN / iSCSI / S2D)**                           | No                                                                                                                           | **No**                                                |
| Minimum SQL edition for two-node design      | Standard                          | Standard (2-node FCI)                                 | **Enterprise**, in practice, for XProtect's multi-database failover (Standard supports only Basic AGs — one database per AG) | **Any supported edition, including Express**          |
| Shared database failure domain between nodes | Yes (replicated)                  | Yes (shared storage)                                  | Yes (synchronized replicas)                                                                                                  | **No (independent databases)**                        |
| Synchronization mechanism                    | Built-in DB replication           | Shared-storage handoff                                | AG replication                                                                                                               | **MIP SDK APIs (application layer) + media backfill** |
| Site independence                            | Limited                           | Limited                                               | Limited                                                                                                                      | **Full**                                              |

### When to Use This Architecture

The cross-subnet RMF architecture is appropriate when any of the following conditions apply:

* The primary and secondary Management Servers must reside in **geographically separate data centers**.
* The primary and secondary Management Servers must reside on **different VLANs or different subnets** within the same data center for network segmentation or security zoning reasons.
* The customer requires **active-active dual recording** with redundant Management Servers, and cannot accept the cost or complexity of a clustered SQL backend.
* The customer requires a **complete, gap-free media archive on both sites** after temporary outages — including outages that occur on the far side of a subnet boundary.
* The customer's DR strategy requires **site isolation** — no shared storage, no shared database, and no shared failure domain between sites.
* The customer's cybersecurity posture requires that **lateral movement** between the primary and secondary sites through database or cluster channels be eliminated by design.
