Servers, virtualization, network and platform operations.
An AI VM with its own GPU or a dedicated machine: compare requirements, resources, maintenance, migration and recovery clearly.
Read article: Local AI with Proxmox: GPU passthrough or a dedicated server?Plan model weights, context and runtime headroom together: assess VRAM classes using a transparent 4-bit calculation and specific GGUF file sizes.
Read article: 8, 12, 16 or 24 GB of graphics memory: which GPU is enough for a local LLM?A specific DL380 Gen11 scenario with NVIDIA L4: configuration, cooling, hardware service and a transparent example budget with net and gross amounts.
Read article: HPE ProLiant as an AI server: what does a professional rack system cost?Start with an existing workstation: calculate components, documented prices and budget reserves for a local AI pilot transparently, with net and gross amounts.
Read article: Local AI for under EUR 2,000: a worked budget for upgrading existing hardwareCPU or GPU, 32 or 64 GB of RAM: distinguish an initial trial, a workstation used regularly and a shared AI service.
Read article: Local AI in business: what is the minimum you need?One fast response is not enough to choose infrastructure. A pilot tests quality, waiting times and operations under representative load.
Read article: Testing local AI: a pilot with representative loadFrom a local test to a shared AI service: plan access, the model API, data flows, backup and operation with Ollama and Open WebUI.
Read article: Ollama and Open WebUI on the company network: a local AI assistantTest a restore with a separate destination, checks by the responsible specialist and a documented result.
Read article: Testing backups: recovery is what proves their valueVirtual machines, Linux containers and separate backups: what matters when using Proxmox VE in a business.
Read article: Proxmox in business: virtualisation with room to chooseCloud, your own servers or hybrid: applications, data flows, total costs and operational responsibility guide the decision.
Read article: Cloud, on-premises or hybrid: comparing operating models sensiblyBackup, local high availability or a second site? A fictional ERP example explains fault domains, RPO/RTO limits and transparent three- and five-year costs.
Read article: Continuity operations and disaster recovery: which level of protection is worthwhile?24 hours of ERP downtime: a fictional business example with a financial impact of 9,150 euros, separate cost categories and explicit three-/five-year scenarios.
Read article: What does 24 hours of IT downtime cost? A calculation example for a mid-sized businessA monitoring alert starts the work. This fictional colocation incident shows how diagnosis, approval, Remote Hands, spare parts and business acceptance determine actual recovery time, with transparent…
Read article: Remote Hands, spare parts and monitoring: how soon is a server running again?Pods, Nodes and the control plane explained: what Kubernetes organizes and which application, data and operational tasks remain.
Read article: Kubernetes explained: what a cluster does for your applicationsYour own server room, Austrian colocation or a hyperscaler: who operates what? A fictional three- and five-year calculation with power, networking, Remote Hands, backup and data protection…
Read article: Austrian colocation, a hyperscaler or your own server room?WireGuard, IPsec and SD-WAN for multiple business sites: assess topology, bandwidth, redundancy, security boundaries and operations using a four-site example.
Read article: WireGuard, IPsec or SD-WAN: which site networking approach fits your operations?Metrics become useful monitoring through responsibility, dependencies and a tested response path. A practical introduction with Zabbix.
Read article: Zabbix: an alert helps when people know who should do whatOne container host or a distributed platform? A decision matrix covering updates, failures, data and operational effort.
Read article: Docker Compose or Kubernetes: which deployment model fits your application?
