Digital engineering studio · Agency
iNTERVOLUTIONS
A Castellón software house that builds and then keeps running: a client fleet, its own SaaS, a database pair and the backups behind both.
- OrbitVPS deployment tool
- Orbit DesktopDesktop client
- TAPEMulti-site editorial engine
- Client platformsWeb, mobile and eCommerce
The problem
- Hard constraints
- One client’s bad day must never become another client’s bad day.
- Root and out-of-band access on everything: they are the operators, not a tenant of one.
- Every site, database and container image restorable from a rehearsed backup, with a retention window they can quote to a client.
- A cost per project that survives a fixed-price quote a year in advance.
- When a client calls them at night, they need to be able to act rather than relay.
iNTERVOLUTIONS is five people in Castellón doing three jobs at once: building web and mobile products, advising on architecture, and then hosting and operating what they shipped. Their own summary of it is "de cero a hero" — from nothing to running — and the third part is the one nobody tenders for and everybody needs.
An agency estate is not one deployment. It is a hundred-odd projects across seven countries, on different stacks and different release rhythms, some of them WordPress and TAPE sites that must simply stay up, some of them their own products, and all of them sharing whatever infrastructure the agency can afford to run well.
What they had was managed cloud, and the bill was the smaller half of the problem. The larger half was that a client incident at eleven at night arrived as a support ticket into somebody else’s queue, and the agency — who the client actually calls — could do nothing but wait and relay.
Architecture
Edge
Edge, DNS and monitoring
3 ×Volt 3
Application
Client site fleet
5 ×Forge 4
Application
SaaS and app platform
3 ×Apex 6
Storage
PostgreSQL pair
2 ×Apex 5
Build
CI, staging and Orbit
2 ×Forge 3
Archive
Backup and retention
2 ×Apex 3
What it runs on
Edge
Edge, DNS and monitoring
3 ×Volt 3
6 vCore · 12 GB · 100 GB NVME
Three small virtual servers doing the unglamorous work: TLS termination and reverse proxying in front of the fleet, authoritative DNS for the domains they manage, and the monitoring that wakes a human. 12 GB is sized for the monitoring, which is the one of the three that has a bad night.
From $18.99 per month
Application
Client site fleet
5 ×Forge 4
i7-8700K · 32 GB DDR4 · 2× 480 GB SSD
Five matched machines carrying the WordPress and TAPE sites they run for clients, split so no single box holds more than a fifth of the estate. Each is a 6-core part at 3.7 GHz with 32 GB and a mirrored SSD pair — deliberately ordinary hardware, because a brochure site that needs exotic silicon is a brochure site that needs rewriting.
From $109.99 per month
Application
SaaS and app platform
3 ×Apex 6
AMD EPYC 7313 · 64 GB DDR4 · 2× 960 GB NVMe
The products they build and operate themselves, on three EPYC 7313 nodes with 64 GB each. Containers land here through Orbit, their own deployment tool, and any one of the three can be drained and rebuilt while the other two carry the traffic.
From $239.99 per month, $230.00 one-time setup
Storage
PostgreSQL pair
2 ×Apex 5
AMD EPYC 7371 · 128 GB DDR4 · 2× 960 GB NVMe
A primary and a hot standby streaming between them, on EPYC 7371 with 128 GB and NVMe. A database is the part of an agency estate that cannot be rebuilt from a repository, so it is the part that gets its own machines and its own failover.
From $239.99 per month, $149.00 one-time setup
Build
CI, staging and Orbit
2 ×Forge 3
Intel Xeon E-2288G · 32 GB DDR4 · 2× 960 GB NVMe
Two runners at 5 GHz for the build matrix, the staging environments and the Orbit release pipeline. Kept off the serving machines on purpose: one client’s Friday deploy must not be able to slow another client’s checkout.
From $109.99 per month, $70.00 one-time setup
Archive
Backup and retention
2 ×Apex 3
AMD Ryzen 7 Pro 3700 · 32 GB DDR4 · 4× 14 TB SATA
Two machines of 56 TB in different regions, holding nightly snapshots of every site, every database and every container image, with a retention window agreed per client. Restores are rehearsed on a schedule rather than assumed to work.
From $219.99 per month, $219.00 one-time setup
Techniques worth naming
A blast radius per client
The site fleet is split across machines rather than stacked on the biggest one available. It costs a little more, and it means an exhausted PHP pool, a runaway plugin or a compromised install takes down a slice of the estate instead of all of it.
Deployment as a product
Orbit is their own tool and every deploy goes through it, their clients’ and their own alike. One path means one place to fix a mistake, and it is the reason a five-person team can operate an estate this size at all.
Rehearsed restores
A backup nobody has restored is a hope. On a fixed cycle a random site and a random database are restored onto the staging machines from the retention copy, and the restore is what proves the backup rather than the backup log.
The database gets its own iron
Everything else in the estate can be rebuilt from a repository and a container registry. The data cannot, so it sits on machines that run nothing else, with a standby that is streamed to continuously rather than restored to on a bad morning.
Staging that predicts production
The CI machines also carry the staging environments, built from the same images and the same Orbit path as production. Different hardware for staging is how a team finds out about a difference at the worst possible moment.
Support
Extended cover, around the clock
Agreed with this account
iNTERVOLUTIONS
Somebody else’s customers. When a shop stops taking orders at midnight the agency is who gets called, so the agency needs somebody to call in turn — and needs it to be a person with root on the machine.
One escalation, any hour
A direct route into the people who administer this fleet, at any hour on any day, without a form and without a first reply that asks them to reboot it.
They speak for their clients
The agency is the account. We do not ask which end customer is affected before we look at a machine — the fleet is theirs, and so is the decision about what matters tonight.
Maintenance around their launches
They publish their launch calendar to us. Platform work is scheduled outside it, and anything that cannot wait is agreed with them before it starts rather than announced after.
Hardware faults are ours
A failed disk or a dark switch port is escalated to the facility by us, with the machine identified, rather than becoming a thing the agency has to chase while their client waits.
Extended cover is agreed per account and written down before it applies. It changes when we are reachable and how quickly a fault is escalated — it does not create a second queue, a premium tier or a response time we have not signed. Everything on the support page is still true here.
What changed
- A client incident reaches somebody with root on the machine, at the hour it happens, instead of being relayed into a queue the agency cannot see.
- A failure is bounded: the estate is spread across machines, so the question after an incident is which slice was affected rather than whether everything was.
- They can put a hosting line in a fixed-price proposal a year out, because the line is a fixed number of machines at a fixed monthly price.
- Restores are something they have done this quarter rather than something they believe would work.
No performance percentages appear on this page. We publish figures we have measured under a named method, and these deployments are described by their architecture instead.
Ready when you are
Bring us a workload, not a shopping list
Tell us what the thing actually does and we will tell you which machines fit — including when the answer is a cheaper one than you asked for.
Shared from $4.99/moVirtual from $6.99/moBare metal from $35.99/mo