What a twin pre-production environment is
In a client's installation the vendor deploys two instances of the software in Docker containers: a production instance and a twin pre-production instance. The machines, network and disks are provided by the client, so the cost of pre-production is in practice the cost of an additional host or of reserved resources on the existing platform.
Twin means that it has the same image versions, the same structural configuration and the same migration chain. The data and the address are different. Thanks to this the update can be run there first, to check whether the start, the migrations and the login pass.
Structure of the cost
The cost of pre-production consists of four parts: hardware or virtual resources, host operating system licences where they apply, the administrator's time for maintenance and the space for backups. Data and backups stay on the client's disks, so the additional disk space must be included in the IT budget.
We do not give software costs here as figures, because pricing is individual. Indicative ranges from 2025-2026 deployments for the whole installation require a conversation about the scope; the price is set individually.
- Compute resources: processor, memory, disk
- Host maintenance by the administrator
- Backup space on the client's disks
- Time to run a trial update
Components of the cost of one incident (sample calculation)
What a failed update costs
The most expensive effect of a failed update is the stop in issuing and closing work permits. Without the system the plant returns to a substitute paper process, and dangerous work usually does not start until the permit is issued. The cost is therefore counted in hours of downtime of the work crews, plus the hours of the administrator and the service team.
Database migrations run on copies of production data of six installations finished successfully, and the run time ranged from 3 to 32 seconds. Such a trial on pre-production is cheaper the closer the pre-production is to production in terms of versions and data.
Sample calculation
We assume cautious figures, which the plant can replace with its own. One failed update per year stops the system for four hours, and during that time ten work crews of three people each are waiting. At a cost of 90 PLN per working hour this gives 4 x 30 x 90 = 10,800 PLN for the downtime alone, without service costs and delays in orders.
Add to this the work of the administrator and the service team during the repair: two people for six hours at 120 PLN per hour, which is 1,440 PLN. Together that is 12,240 PLN per incident. Pre-production does not eliminate the risk completely, but it lets you catch migration and configuration errors before production.
When pre-production pays off
The return depends on how often incidents happen and on how many people are stopped. In a plant with several dozen work crews on a shift, one incident a year approaches the annual cost of keeping a test host. In a plant with a few crews the threshold is higher, and a lighter pre-production on shared resources is enough.
That is why the purchase question is worth phrasing differently: how many hours of plant downtime correspond to the annual cost of pre-production. If the answer is one or two hours, the additional machine is justified without further analysis.
Cost of one failed deployment
| Parameter | Value |
|---|---|
| Downtime | 4 hours |
| Work crews waiting | 10 crews |
| People per crew | 3 people |
| Cost of one working hour | 90 PLN |
| Repair: 2 people, 6 hours, 120 PLN | 1440 PLN |
Result: 10,800 PLN of downtime plus 1,440 PLN of repair gives 12,240 PLN per incident.
Note: Sample calculation based on the assumptions above. Every plant enters its own figures.
Video
Frequently asked questions
Does pre-production have to be a separate machine?
It does not have to be, but it should have the same image versions and configuration as production. It can run on the shared resources of the client's platform.
Who pays for the pre-production hardware?
The infrastructure is provided by the client. The vendor deploys the software in Docker containers on it.
Does pre-production guarantee there will be no failures?
No, it does not guarantee that. It reduces the risk, because migration and configuration errors surface before the deployment to production.
How can a plant calculate its break-even point?
Enter the number of work crews, the hours of downtime and the cost of an hour of work. Compare the result with the annual cost of keeping a test host.
Let's talk about the installation in your server room
We will present the requirements and the split of responsibilities for your plant based on the installation documentation. Book a conversation with the InnerWeb team.
Let's talk