Why pay for infrastructure that you do not need? If you host with one of the big providers, you will be paying outlandish bills. And you probably do not need them. You should not automatically inherit the cost, complexity and dependency of systems designed for global hyperscale.
Design the workload properly before buying more infrastructure. Cloud-capable, but not cloud-captive.
A micro-server is a compact, single-board or small-form-factor machine running a complete application stack: web application, relational database, task queue, cache and object storage. It is not a toy or a demonstration; it is a production server sized to the actual workload.
At kawalek.uk, we design and build our own micro-servers. The hardware is chosen for the workload: real-world efficiency, maintainability and a clear upgrade path rather than marketing specifications.
Micro-server deployment is our primary design target, not merely one option in a list. The application is engineered to run on compact infrastructure from the start.
A poorly designed application creates unnecessary demand on infrastructure. Fixing the design is cheaper than buying more hardware.
Infrastructure cost is wider than the advertised price of a virtual machine. A complete managed deployment may include some or all of the following, billed separately.
Not every organisation pays all of these charges. Total cost of ownership requires examining the full stack, not just the headline compute price.
01
Django, async-capable WSGI/ASGI server. Session, cache and queue clients configured from start.
02
PostgreSQL for structured records, transactional state and relational queries. Properly indexed from day one.
03
Celery with Redis. Expensive work deferred. Cache used aggressively to keep requests fast.
04
S3-compatible local storage for binary assets, exports and backups. Cloud S3 available as a drop-in extension.
A micro-server requires active management: software updates, backups, monitoring, and occasional hardware maintenance. kawalek.uk applications include structured logging, health-check endpoints and automated backup jobs from the start.
The application stack uses established open-source components — Django, PostgreSQL, Redis, Celery — with well-documented operational practices and no proprietary lock-in.
Single-node deployment is appropriate for many business workloads. Where higher availability is required, options include:
A micro-server-native application is not locked to a single node. Because the stack uses standard interfaces and no proprietary managed services, the workload can be migrated at any point:
Migration is a strategic choice, not a technical crisis.
Organised by deployment model — not as a balanced menu of equal options.
| Deployment model | When it fits | When to consider moving |
|---|---|---|
| Micro-server native | Stable internal workload, predictable traffic, local control required, sensitivity to egress cost. | Concurrent user count grows rapidly; staff cannot support the server. |
| Micro-server with redundancy | Business-critical operations; uptime SLA required; backup site needed. | Replication complexity exceeds benefit; cloud cost becomes lower. |
| Hybrid deployment | Core application on micro-server; AI, email or object storage on selective cloud services. | Cloud dependency grows to the point where micro-server offers no advantage. |
| Public-cloud-led | Highly variable public traffic; global distribution; specialist managed services essential. | N/A — appropriate choice for this workload. |
Micro-servers are not suitable for every workload. We will say so clearly:
kawalek.uk designs and builds its own micro-servers. Hardware and software are co-designed: components are chosen for real-world efficiency and long life rather than marketing specifications. Clients receive a fully integrated, maintainable stack with a clear upgrade path.
The European Commission's Data Act identifies high egress charges, lengthy switching procedures and insufficient interoperability as cloud switching barriers. Its provisions require providers to facilitate data export and switching, with certain switching-related charges scheduled for removal from 12 January 2027.
Cost comparisons between micro-server and cloud deployment depend on workload, team capability, uptime requirements and specific service choices. We do not claim universal savings. Every engagement begins with a workload assessment.