General purpose compute tuned for your specific needs
Readily available industrial servers and edge computers from multiple vendors, pre-loaded with an open stack plus the software you require. Configured, tested, and documented.
Three nodes or more
True redundancy via three or more nodes. The diagram shows a typical configuration, with redundant networking and software defined storage. Add nodes for more capacity.
From the first conversation to a running system
There is no cart and no list price, because a cluster is configured to your requirements. This is how it goes.
- Scope
Tell us about the plant
Tell us what you run, what has to survive a failure, and what it has to talk to. From that we specify nodes, storage, and networking. If the requirements are not settled, an assessment comes first.
- Build
Quoted to order
Hardware, licensing, and configuration in one price. It ships with a scripted node build, a failure-injection test plan and its results, as-built documentation, and an operations guide.
- Run
Support once it runs
Priced per machine and split by class, so a three-node cluster is three machines. Combines with software support when we also look after what runs on it.
Common example deployments
Ranges from delivered clusters, complete: hardware, licensing, and the configuration work that makes it a tested system. Every cluster is quoted to order, but the bands below give you an idea of what to expect.
Common cluster pricing
Edge
1 node$ 16,000 to $ 40,000
A single node, standard or server class, with its rack, UPS and network switch. The bottom of the band is a compact standard-class machine, a panel PC or DIN-rail IPC. The top is a server-class node wired so nodes two and three can drop in later without a rebuild.
Plant
3 - 5 nodes$ 65,000 to $ 135,000
No single points of failure. The spread is node class and operating system licensing, with our labor staying relatively constant. The bottom of the band is a three-node standard-class Linux/Ceph cluster; the top is a three-node server-class Windows S2D cluster.
Enterprise
5+ nodesQuoted
Many nodes, many sites, or complex deployments.
What affects pricing
- Hardware is the main cost driver The band is wide because the hardware can be a compact standard-class node or a server-class node. The labor to configure and validate the cluster is relatively constant across the band, so the hardware choice dominates the price.
- Node class A desktop-class mini PC against a rackmount server with ECC memory, redundant power, and a 3-year warranty. Roughly 3x on hardware per node, and the single biggest swing in any quote.
- Operating system licensing Windows Server Datacenter runs about $ 5,900 per node, and the Linux stack we prefer costs nothing to license. We also do Windows guests on Linux hosts.
- Storage How many NVMe drives per node, how much you need to keep and whether your workload is write-intensive. Endurance is priced as well as capacity.
- Network fabric Shared 10G uplinks, or a dedicated 25G mesh carrying storage replication.
- Platform scope Do you need virtualization, a database, directory services, or monitoring and alerting? Each adds a fixed cost on top of the cluster, but we only configure what you need.
Configuration Options
- Kubernetes platform, $ 6,000 to $ 10,000 Kubernetes with replicated storage and a high-availability database operator, ready for containerized workloads. The lower end is a single node; the upper end is highly available across the cluster.
- Commercial database, $ 19,000 to $ 25,000 SQL Server licensing plus a deployed and configured database platform, when the application requires it. Licensing is per core and can rival the hardware. The upper end adds a second replica for failover.
- Directory and platform services, $ 7,500 Active Directory, DNS, DHCP, update management, and a group policy baseline, deployed on the cluster.
- Windows Server licensing, $ 5,900 / node Windows Server Datacenter, when the platform is Windows or guests need it. One license per node covers unlimited Windows Server virtual machines on that node. A line item rather than a kind of cluster, because a Linux host running Windows guests is a normal build.
- Monitoring and alerting, $ 5,000 to $ 10,000 Prometheus, Grafana, and Alertmanager with dashboards, alert routing, and a runbook behind every alert. The lower end is a single node; the upper end is a cluster. Prerequisite for Proactive Operations.
Terms
- Each cluster is individually quoted based on the hardware and software you need.
- If the scope grows mid-project, we price the change and get your approval before doing the work. No surprise invoices.
- All hardware and software is billed up front. Configuration services are 60% up front and 40% on delivery with operations guide and validation documentation.
Tested and documented
Most clusters are configured by hand and never tested against a failure. We build infrastructure the way we build software: scripted, tested by injecting failures, and documented with the results.
- A scripted, repeatable node build, so you can reinstall a failed node or add one without us
- A failure-injection test plan and its results, run before the system ships
- As-built and network design documentation, plus a credential inventory
- An operations guide written so anyone can operate the system, not just the engineer who built it
Sanitized examples from a delivered four-node cluster.
Hardware support rates
Priority Response commits to a 4-business-hour answer and includes the labor to put the system back the way we delivered it; Proactive Operations means we watch and act before you call. Priced per machine, split by class. A three-node cluster is three machines.
| Machine class | Priority Response | Proactive Operations |
|---|---|---|
| Standard hardwareMini PCs, DIN-rail IPCs, panel PCs | $ 100 / machine / mo | $ 200 / machine / mo |
| Server classECC memory, redundant power, or out-of-band management | $ 250 / machine / mo | $ 500 / machine / mo |
A plan includes the labor to bring a covered machine back to its baseline: the configuration we delivered, plus any improvement you have paid for since. That work is never billed by the hour under a plan. Anything else is new work, scoped and quoted as its own project: new capability, changes you ask for, and equipment or software we did not deliver. Changes made outside our change process take the affected component out of coverage until it is re-validated, which is quoted the same way. Hardware faults go to the manufacturer warranty; we coordinate the RMA. If the labor a plan absorbs runs past its fee, the rate moves at renewal, never mid-term.
Support is $ 350/hr with a 2-hour minimum and a 5 business day response time when you have no plan. Commitments are business hours; after-hours work bills at the overtime rate. Stackable coverage is below.
Add-ons for hardware
Optional, per covered deployment, on any response tier.
- Annual Re-Validation, $ 5,000 / yr. We re-run the full failure-injection test campaign each year and deliver a written report. Included in Proactive Operations.
- Annual Stack Upgrade, from $ 8,000 / yr. One major-version upgrade a year across the platform layer, tested and cut over in a maintenance window. Included in Proactive Operations.
- Monitoring & Alerting Enablement, $ 5,000 to $ 10,000 one-time. Prometheus, Grafana, and Alertmanager deployed with dashboards, alert routing, and runbook links. Prerequisite for Proactive Operations.
Licenses and hardware we procure
Provided under any project at cost plus a 20% margin, itemized as its own line, the same way our bills of materials show it. The margin is identical whoever we buy from, so nothing about what we specify depends on which vendor discounts hardest. We pick on fit.
Applies to anything on a bill of materials, from PLC hardware to SCADA licenses.
Tell us about your plant
There is no cart and no list price. Clusters are configured and quoted to your requirements. Send the details and we'll follow up.