Objlab
← Back to products

Objlab Appliance

See pricing →

For Cerbercode, GISMart and ObjChain there is always the service version, and that's what we recommend. The appliance comes into play when the customer cannot use it: regulated sectors, data that by policy cannot leave the network, tenders that require installation on-site. In that case, the problem isn't the software — it's already ready to run on-premise — but everything around it: scaling, the operating system, container runtime, certificate, backups, updates. This is the box in which all of that work has already been done.

What it solves

  • Policies or tender documents that prohibit the data outside the corporate network
  • No one at home who wants to maintain a container runtime
  • Size a server for a workload that is not yet known
  • Updates that become a project instead of an operation

What is it

A 1U server rack with the operating system, container runtime, and an Objlab stack — Cerbercode, GISMart, or ObjChain — already installed and tested. It arrives pre-configured: simply mount it in the rack, connect it, and turn it on.

When it makes sense, and when it doesn't

Makes sense when the data cannot leave the customer's network: healthcare, defense, public tenders, groups with strict policies. In those cases, the service version is not an option, and the alternative to the appliance is a customer iron installation project — slower and more fragile.

It doesn't make sense if the service version is practical. It costs less to the customer, updates itself automatically, and does not force us to send someone when a disk fails. We say it before instead of after: an appliance sold to those who didn't need it becomes our maintenance problem for five years.

Technical specification
Format 1U rack, standard depth
Processor server class, 8-16 cores
Memory 128 GB ECC
System disk dedicated, separate from data
Storage 2 × 2 TB in RAID 1, approximately 2 TB usable
Network 4 ports
Power supply dual, redundant
Management IPMI, independent remote console

ECC Memory. Corrects single-bit errors on its own. On a machine that holds customer data and runs unsupervised for years, the difference between ECC and non-ECC is between an error that is corrected and logged and one that silently becomes incorrect data.

Separate system disk from data. The operating system and containers reside on their own disk. Thus, a data recovery does not affect the system, a system update does not put the data at risk, and the two fill up independently: an insane log does not stall the application.

Four network ports. Because in an appliance they don't all do the same thing: the customer service network, the backup network, out-of-band management, and a spare one. Making them all go through a single port means that when it fails, everything goes down together.

Dual power supply and an IPMI console complete the picture: these are what distinguish a server from a racked PC, and they are not removed to lower the cost estimate.

Let's be upfront and say why: the cost of server memory has increased by nearly five times in a year, and the shortage is expected to continue at least until 2027. A price list published today would already be wrong in two months.

A quotation for this configuration will be provided today. It includes hardware, installation of the stack, and testing; product licenses are quoted separately, according to their price list.

The installation of the stack, the pre-shipping testing, the configured backups and the update channel.

And the sizing. This configuration covers the vast majority of cases, but for large loads measure first instead of guessing: an undersized appliance is discovered in production, which is the worst time to find out. If more memory or space is needed, it should be said during the estimate phase, not after.

Compliance & regulations

Frequently asked questions

Perché sulla scheda non c'è il prezzo?

Today, a published price list would last only a few weeks. The cost of server memory has increased by almost five times in twelve months -- producers have shifted capacity to memory for artificial intelligence -- and the shortage is expected to continue at least until 2027. A number written today would be underpriced in two months, and at that point we either keep our word or sell at a loss. We prefer to give a real quote, on the day.

Mi conviene l'appliance o la versione a servizio?

Nella maggior parte dei casi la versione a servizio, e lo diciamo prima invece che dopo: costa meno, si aggiorna da sola e non obbliga nessuno a intervenire quando un disco cede. L'Objlab Appliance ha senso quando il dato non può uscire dalla vostra rete -- sanità, difesa, capitolati pubblici, policy di gruppo stringenti -- cioè quando la versione a servizio non è un'opzione.

Che cosa comprende il preventivo?

Hardware, installation of the Objlab stack, pre-shipping testing, configured backups and update channel. Product licenses are separate, according to their price list. VAT and shipping excluded.

E se ci servisse più di 128 GB di memoria?

It should be said at the beginning of the negotiation, not at the end: the reference configuration is at the top of its platform, so beyond 128 GB we don't add memory but change machine class, with a different estimate. For large loads we measure first instead of guessing -- an undersized appliance is discovered in production, which is the worst time.

Posso far girare più prodotti Objlab sulla stessa macchina?

Sì, ed è spesso la scelta giusta. Quanti ne regga dipende dal carico reale, non dal numero: è parte del dimensionamento che facciamo in fase di preventivo, insieme alla scelta di memoria e dischi.

Chi assiste l'hardware se si rompe qualcosa?

We. The appliance has ECC memory, redundant dual power supply, and an operating system-independent management console: these are the three things that allow us to diagnose and replace without stopping everything, and we don't remove them to lower the quote.

Parla con un esperto