Reliability
When a node fails, the work moves.
Inside your private cloud first, then across a network of 1,800+ nodes if you allow it. The uptime commitment comes with the tier you choose.
Failover
Three rules you set once.
Inside your private cloud first
Work restarts on another node you own or rent inside your private network.
Then across the network
If you allow it, across 1,800+ nodes behind your private cloud, under the same placement policy.[1]
Never past your policy
Work pinned to named devices fails over only between the devices you named.
Uptime commitments
By tier, in a 30-day month.
| Tier | Commitment | About | Fits |
|---|---|---|---|
| A, Regulated | 99.99% | 4 minutes down | Regulated data, public sector |
| B, Enterprise | 99.95% | 22 minutes | Most production work |
| C, Edge | 99.9% | 43 minutes | Latency-bound work |
| D, Public | 99.5% | 3 hours 36 minutes | Restartable work |
The figures are commitments in the master services agreement, with the credit mechanism and the claim window in the same agreement. Nobody credits you automatically, so file the claim.[2]
What protects your data
Backups and restore, by product.
Managed databases
Three-node high availability with automatic failover. Continuous archiving holds seven days of history; restore to any second, into a new database, while the source keeps running. Backups are kept in a separate region.[3]
Virtual machines
Snapshots and backups over the API and the console. A snapshot takes storage close to the size of the instance disk.[4]
Kubernetes
Managed control plane. Keep state in managed databases or object storage; a second cluster plus DNS gives a second region.
Test it yourself
Launch two nodes, stop one on purpose, and watch the work come back on the other. The console shows where it landed.
Get started
Pull the plug on a node and watch.
The free tier is enough to test failover in an afternoon.
- 01
Join the waitlist.
Tell us what you want to run and where. No credit card.
TakesTwo minutes
- 02
Launch a machine or a cluster.
From the console, the API or the CLI.
Takes2 to 5 minutes
- 03
Bring your own servers.
Triple X sets up your private cloud on them.
WhoTriple X
We never ask for your root credentials, API keys or bank details by email.
What to ask any provider
What to ask before the first outage.
Seven questions, and the document that should answer each one.
| Ask | Triple X Cloud | Where to verify |
|---|---|---|
| Where does work move when a node fails? | Inside your private cloud first, then the network if allowed | Reliability page, console placement policy |
| Can work leave devices I named? | No, unless your policy allows | Console placement policy |
| What is the uptime commitment? | 99.5% to 99.99% by tier | Tier table; master services agreement |
| Are credits automatic? | No; file the claim | Master services agreement |
| Database restore window? | Seven days, to any second, separate region | Database documentation, August 1, 2026 |
| Machine disk protection? | Snapshots and backups over the API | Virtual machine documentation |
| Can I test it free? | Yes, two nodes on the free tier | Published pricing, free tier |
Questions
What buyers ask before they sign.
In the order they come up. Every answer ends with where to check it.
10 answers
All questionsWhat happens when a node fails?
The work moves. Failover follows your private cloud's policy: inside your private cloud first, onto another node you own or rent, then across the 1,800+ node network, a company-stated figure, if your policy allows it. The console shows where the work landed.
Proofthe three failover rules on this page and the company-stated fleet count, October 2026.[1]
Will my work ever leave the devices I named?
Not unless you allow it. Work pinned to named devices fails over only between those devices. Failover follows your private cloud's policy, and the policy is yours to set once and change in the console.
Proofthe placement policy screen in the console and the third failover rule on this page.
What uptime do you commit to?
By tier, in a 30-day month: Tier A Regulated 99.99%, Tier B Enterprise 99.95%, Tier C Edge 99.9%, Tier D Public 99.5%. The commitment, the credit mechanism and the claim window sit in the master services agreement.
Proofthe tier table, August 1, 2026, and the service levels in your master services agreement.[2]
How do I claim a credit?
File it. Nobody credits you automatically. The claim window and the credit mechanism are in the master services agreement, so read both before an outage, not after. Keep your own monitoring record to attach to the claim.
Proofthe credit and claim clauses in your master services agreement.
Does failover cost extra?
The published rate card has no failover line. You pay the hourly rate for the capacity your work runs on, wherever it lands inside your policy. Tier A to Tier C capacity is priced on quote; Tier D uses the published rates.
Proofthe published rate card and the tier price index, August 1, 2026.[2]
Does a failover expose my data?
No. Failover follows your private cloud's policy, and the replacement node joins the same private network under the same placement rules. The platform underneath still sees only that a workload exists, its size and its cost, not what is in it.
Proofthe access table on the Security page and the placement policy screen in the console.
How are databases protected?
Three-node high availability with automatic failover. Continuous archiving holds seven days of history; restore to any second into a new database while the source keeps running. Backups are kept in a separate region. Private endpoint by default.
Proofthe managed database overview in the documentation, August 1, 2026.[3]
Are machine disks backed up for me?
Snapshots and backups run over the API and the console, on a schedule you set. A snapshot takes storage close to the size of the instance disk. Local NVMe lives with the machine, so set the schedule before you need it.
Proofthe snapshot and backup endpoints in the virtual machine documentation, August 1, 2026.[4]
How do I test failover myself?
Launch two nodes on the free tier, stop one on purpose, and watch the work return on the other. The console shows where it landed. An afternoon is enough, and it costs nothing on the free tier.
Proofthe free tier in the published pricing and the placement view in your console.[5]
What is not covered?
Application state you did not put in managed databases or object storage. A second region for Kubernetes, which needs a second cluster plus DNS that you set up. Outages inside your own site: power, racks and your network stay your responsibility.
Proofthe backup notes by product on this page and the responsibility split in your service agreement.[6]
No questions under this term yet. Ask Triple X.