SecurivePlatform
One data path, from a running workload to a proven restore.
Backup, replication, orchestration and verification are one product with one data model — not four integrations behaving like a platform. This page is the how, including the parts we don't do.
The data path
Five stages, and what each one guarantees.
Capture
Read only what changed, without pausing the workload.
- Changed-block tracking means a nightly incremental reads deltas, not disks.
- Snapshots are taken through the hypervisor, so the guest keeps running.
- Where a guest agent is available, writes are quiesced so databases land application-consistent rather than crash-consistent.
Transport
Compress and encrypt before anything leaves your network.
- Deduplication and compression run at the source, so the link carries the smallest honest payload.
- Encryption happens before transmission with the tenant's own key — the transport never sees plaintext.
- Connections are outbound only. No inbound firewall rule, no port forwarded to a backup server.
Store
Write once, into a bucket in the region you chose.
- Storage is versioned: a delete hides a restore point, it does not destroy one.
- Retention governs deletion — the console has no button that shortens it.
- Each tenant gets its own datastore, its own key and a hard quota, so one tenant can neither read nor starve another.
Replicate
Keep a second copy somewhere the first disaster can't reach.
- Replication is continuous rather than a second scheduled job, so a lost site costs minutes rather than a day.
- The second region is one you choose, and data is never relocated for capacity reasons.
- A replica of an encrypted volume is an encrypted volume — replication alone is not ransomware protection. What makes it one is the versioned history behind it, which the backup credentials cannot erase.
Verify
Restore it, boot it, check it, and write down how long it took.
- Restores run on a schedule into an isolated network with no route to production.
- The workload is booted, the filesystem checked and the application probed — a job can report success and still produce an image that will not start.
- The measured RTO goes into a report you can hand to an auditor, a customer, or an insurer.
Architecture
What runs where, and what we can see.
Two questions decide most security reviews: does our data pass through you, and what do we have to open. The answers are no, and nothing.
- Control plane
- Hosted by us. Holds policies, schedules, job history and metadata — never your backup content in plaintext.
- Data plane
- Your data goes from your infrastructure to object storage in your chosen region. It does not transit our console.
- Connectivity
- Outbound HTTPS only. Nothing needs to be reachable from the internet, and no inbound rule is required.
- Agents
- None for hypervisor backup. Physical servers and application-consistent database backup use a lightweight agent.
- Keys
- One AES-256 key per tenant, wrapped by a master key held outside the database.
- Storage
- S3-compatible object storage, versioned. Enterprise plans can point at their own bucket.
Recovery
Six ways back, because incidents aren't all the same size.
Restoring a deleted spreadsheet and recovering a failed region are different problems. Using the same mechanism for both makes one of them slow.
- File and folder
Someone deleted the wrong thing.
Browse a restore point like a filesystem and pull back one file. No full restore, no staging space, no ticket to the storage team.
- Seconds
- Whole machine
A VM is gone and the business is waiting.
The machine is rebuilt on the hypervisor from its restore point. How long that takes is set by how much data has to travel, so the honest answer is a rate, not a number.
- Proportional to its disks
- Point in time (databases)
A bad migration ran at 14:07.
Replay to a chosen moment using base backups plus shipped transaction logs — recovery to 14:06, not to last night.
- Minutes to hours
- Bare metal
The hardware itself failed.
Restore a physical server to different hardware, with drivers injected so it boots on a machine it was never installed on.
- Hours
- Granular Microsoft 365
A mailbox, site or Teams channel is missing.
Restore a single mail item, a SharePoint document version or a whole site, in place or to a different target.
- Minutes
- Orchestrated failover
The site or region is down.
Bring a group of workloads up in the second region in dependency order, with networks remapped — then fail back cleanly when the primary returns.
- Minutes, per runbook
Constraints
What this platform doesn't do.
You'd find every one of these during a proof of concept. Finding them here costs you less.
- Application-consistent backup requires a guest agent. Without one, backups are crash-consistent — fine for most filesystems, not recommended for databases.
- RPO is bounded by bandwidth. A workload changing faster than the link will fall behind, and the dashboard reports that honestly instead of showing green.
- There is no tape support, and no S3 object lock — it cannot coexist with a deduplicating chunk store, which is what makes the storage efficient. Deletion protection comes from versioning and a separated delete permission instead.
- We do not import another vendor's backup format. Migration means running in parallel and proving a restore, not converting a history you have never tested.
- Microsoft's Graph API throttles. The first full backup of a large tenant takes days — a platform limit rather than a product one.
By workload
The specifics for what you actually run
Microsoft 365
Exchange, SharePoint, OneDrive and Teams, restored item by item.
DetailsServers & databases
Physical servers, SQL, Postgres and MySQL to a point in time.
DetailsReplication & DR
Minutes of RPO, one-click failover, and scheduled drills.
DetailsFor MSPs
Multi-tenant console, white label, and per-tenant billing.
DetailsQuestions
The things people ask before they switch
No. The control plane holds policies, schedules and job metadata. Backup content goes from your infrastructure to object storage in the region you chose, encrypted with your tenant key before it leaves your network.
Nothing inbound. Everything is outbound HTTPS, so there is no port forwarded to a backup server and nothing of yours is exposed to the internet.
Not for hypervisor backup — that is agentless. Physical servers and application-consistent database backup use a lightweight agent, because there is no other honest way to quiesce a database from outside the guest.
After the first full backup, only changed blocks move, compressed and deduplicated. The first full is the expensive one, and it can be seeded or throttled to a window so it doesn't take the office link with it.
The job resumes from where it stopped rather than starting over. A partially uploaded restore point is never presented as complete.
Yes. Everything the console does is available over the API, with scoped keys and webhooks for job events, so backup policy can live in the same pipeline as the rest of your infrastructure.
Find out whether your backups actually restore.
Connect one hypervisor, set one policy, and let a scheduled drill try to bring it back. If it doesn't, you'll know in a day rather than during an incident.
No credit card · 14-day trial · Migration assistance included