Explanations that make sense.
Articles about networking, colocation and Odoo — in plain language, written by our engineers. No marketing stories, just answers.
A GPU server in colocation: power, cooling and what fits in a rack
A rack has 42 units, but that is not the limit for GPU servers. A worked example with three-phase 32 A per feed, what A+B redundancy does to your budget, why heat is the real limit and what you need to supply up front.
Classless delegation in practice: setting up, testing and troubleshooting
Part 2 on RFC 2317: how to set up the delegated reverse zone on your side, how to check the whole chain with dig, and the mistakes we see most often in practice.
Classless reverse DNS delegation: how RFC 2317 works
Reverse DNS can only be delegated per octet, so a block smaller than a /24 has no zone of its own. RFC 2317 solves that with a subzone and a CNAME per address. Part 1: the mechanism.
Failover with BGP: private ASN, VRRP and AS-path prepending
A redundant connection does not have to be complicated. How to build a failover you understand yourself with a private ASN, VRRP, AS-path prepending and BFD, and how our own anycast load balancers fail over.
Managed vs unmanaged VPS: what managed really covers, layer by layer
Managed VPS means something different at every provider. Four layers (infrastructure, OS, stack and application) show what you are buying, who manages what, and whether you keep root.
Blackholing during a DDoS attack: the emergency brake, and how to use it
A blackhole makes one address unreachable so the rest of your network stays reachable. Three ways to use it with us: a switch in the dashboard, your own BGP session, and automatically for the edge addresses of your subnets.
DDoS notifications in Slack, Discord or your own system
An attack notice in your mailbox is fine, but your team lives in a chat channel and your monitoring wants something a machine can read. How you receive DDoS notifications as a webhook, and how you check that a message really comes from us.
NIS2 supplier requirements: what to ask your network and hosting provider
Under NIS2 you have to show that you manage the risks of your suppliers. Which questions to ask your network and hosting provider, which evidence counts, and what we answer ourselves.
Keeping an eye on your commit: transit traffic during the month
You only notice going over your commit on the invoice, unless you see it coming. How you follow your 95th percentile during the period in our dashboard, when we warn you, and why an interim figure early in the month misleads.
How to get your own ASN: RIPE requirements, costs and lead time
What RIPE NCC asks before it assigns you an AS number: multihoming, a routing policy of your own and a contract. Plus what it costs, how long it takes and what you still need to sort out afterwards.
Setting up reverse DNS: why your mail server needs a PTR record
A PTR record ties an IP address to a name, and without it mail is often refused. Why you do not set it at your DNS provider, and how you arrange it per address in our dashboard.
Cybersecurity Month 2026: six jobs for your infrastructure
October is European Cybersecurity Month. Not a campaign pitch but a short list to work through as the person responsible for IT: admin accounts, patch status, DMARC, management access, your incident contact and a restore test.
VMware to Proxmox migration: costs, pitfalls and a step-by-step plan
Proxmox has no per-core licence, but a migration from VMware is not free. Where the real costs of a Proxmox migration sit, which pitfalls you can see coming, and the order to do it in.
Giving colleagues access without showing them everything
One shared login for the whole team feels practical, until someone leaves. How you invite colleagues in our dashboard with exactly the rights they need, and why a contact and a user are two different things.
Switching to an OpenAI-compatible API: what actually changes
A different base_url and a different model name, and your application talks to an open model. That is true for the first call. Where compatibility ends, what you need to recalibrate, and why you should build the way back in from the start.
Following a DDoS attack: what you see and what happens
What happens between the first traffic spike and the notice that an attack is over, and how you follow each attack as a customer in our dashboard: target, vectors, what reached you and what we did.
RAG without your documents going out over the internet
With RAG it is not just your question that leaves the building: every call carries a piece of your documents. What goes over the line at each step, what re-indexing costs in bandwidth, and when keeping everything in one data centre actually pays off.
Reporting the energy and CO2 of your IT: scope 3, two methods and what to ask your provider
Colocation and AI inference are scope 3 for you, not scope 2. What Omnibus I and VSME mean for anyone still reporting, how the location- and market-based methods work with the 0.244 factor, which figures to demand from your provider, and why grams of CO2 per request is false precision.
The CLOUD Act and your AI provider: what sovereignty does and does not mean
An EU region is not EU jurisdiction. How the CLOUD Act works through ownership, why the Data Privacy Framework is under pressure, what even a Dutch provider cannot promise, and which questions to ask before you send prompts with customer data to an API.
Tokens explained: what an LLM call really costs
A language model does not count in words but in tokens, and you pay for everything you send along, not just your question. A worked example with a Dutch prompt, why Dutch costs more than English, and when dedicated becomes cheaper than per token.
Your rack from a distance: what is in it, what it uses and how you get in
Colocation means your hardware is where you are not. How our dashboard shows what is in your rack, what each unit consumes and how you arrange access to the data centre.
Choosing open models: licence, origin and Dutch
Open-weight is not the same as open source. What the licences of Mistral, Gemma, gpt-oss, EuroLLM, Qwen and Llama actually allow, how well they handle Dutch, and how to work out whether a model fits in memory.
What is AI inference, and why it matters where it runs
Training and inference are two different things, and only the second touches your day-to-day data. Why every call is a data transfer, what TLS does and does not solve, and what the model's location determines.
When custom software is the wrong answer
From a software builder you expect the advice to build. The honest answer is often no: when standard software or configuration wins, and where custom development is actually worth its money.
Monitoring is not the same as error tracking
Monitoring tells you whether the service is up, error tracking tells you what went wrong in the code while all the lights were green. Why you need both and which tool answers which question.
CI/CD without buzzwords: why a release is not an event
Deploying often sounds risky, but the opposite is true: the bigger and rarer the release, the harder the blow. How our pipeline works and what you notice as a client.
"Where will this run?" should be a late question
In many software projects the hosting choice steers the build. With declaratively delivered software it is a late, cheap choice: SaaS with us, on-premises with you or a bare machine — same project.
Network configuration under version control: every change a commit
The configuration of a switch or router is code, but rarely treated that way. Why we automatically record every configuration change in git, and what that delivers during outages and audits.
Out-of-band management via WireGuard: your BMC does not belong on the public internet
Management interfaces like iDRAC and iLO give full control over your server and should never sit behind a public IP. How our separate out-of-band network with WireGuard access works.
Day one on an existing project: time-to-first-commit as a quality metric
How long before a new developer can run your project? Why that question is about risk rather than comfort, and what we deliver to keep the answer short.
Seeing errors before your customer calls: error tracking with GlitchTip
Users rarely report errors, they click away. How error tracking works, why we run GlitchTip self-hosted and what that means for privacy and the GDPR.
NIS2 in practice: how we handle the Dutch Cyber Security Act
On 15 August 2026 the Dutch Cyber Security Act enters into force — the national implementation of NIS2. What the law requires (duty of care, reporting, registration) and how we set it up ourselves.
Bank reconciliation in Odoo 19: the new Bank Matching view
Odoo 19 replaces the reconciliation widget with the Bank Matching view: suggestions per line, matching that learns from your choices, chatter on transactions and reconcilable draft entries.
Building your own cloud: the architecture behind PlanetNode
How we built a complete open source cloud platform on Proxmox for our sister company PlanetNode: our own template pipeline, our own metadata service and provisioning driven straight from the customer portal.
95th percentile billing explained
How 95th percentile billing works: a data point every five minutes, the highest 5% discarded. Why short peaks cost you nothing and how to pick your commit.
Odoo as the engine behind your webshop: the FiberNXT case
How the FiberNXT B.V. webshop (Medusa) integrates with Odoo: webhooks, real-time stock, automatic sales orders and payments that reconcile themselves — and where to draw the line between shop and ERP.
Data centre migration: the runbook
A data centre migration is not a move of equipment but of dependencies. How you work out the boot order and what has to be in place before the first server leaves the building.
Self-hosted, Odoo.sh or Odoo Online: which do you choose?
The three ways of running Odoo compared honestly: Odoo Online, Odoo.sh and self-hosted. Costs, customisation, upgrades and data residency.
What is colocation?
Colocation explained: your own servers in a professional data centre. What you get, how it differs from cloud and an in-house server room, and when it fits.
DDoS attacks: how to protect your infrastructure
How DDoS attacks work — volumetric, protocol and application layer — and the defence that goes with each: from BCP38 and blackholing to FlowSpec and scrubbing (NaWas).
EVPN-VXLAN explained: Layer 2 without the classic misery
How EVPN-VXLAN works: VXLAN as the dataplane (L2 over a routed network), EVPN as the BGP control plane. And why this has replaced classic L2-over-WAN with spanning tree.
OSS VAT in Odoo: EU sales without foreign registrations
Do you sell to consumers in other EU countries? Here is how to set up the One Stop Shop (Union scheme) in Odoo: fiscal positions, rates and the report for your return.
What does colocation cost per rack?
How a colocation price is built up: space, power and network as three separate items, plus one-off costs. With a worked example and starting prices.
BGP communities: tagging routes and steering traffic
What BGP communities are and what you can do with them: informational tags, traffic engineering with prepends, selective announcements and blackholing. Including the difference between standard, extended and large communities.
Power in the data centre: A/B feeds and what you actually use
Two power supplies does not mean twice the power. How A and B feeds work, why your connected load and your consumption are two different things, and what you are billed for.
Odoo 19 and the end of XML-RPC: migrating to JSON-2
Odoo 19 deprecates XML-RPC and JSON-RPC; the replacement is the JSON-2 API. What does that mean for your integrations — and how do you migrate in time?
What is a wavelength (DWDM)?
Wavelengths explained with a simple analogy: one optical fibre, dozens of colours of light, and one colour entirely for you. On DWDM and when a wave fits.
Regex in reconciliation models
Regex in Odoo reconciliation models: Match Regex and From label. Examples and a testing tip. In plain language.
BGP transit or Managed Internet Access?
The difference between internet over BGP/IP transit with your own AS and Managed Internet Access (MIA). When do you pick which — and what does it mean for control, redundancy and growth?
Reconciliation models for automatic bank reconciliation
Reconciliation models in Odoo explained: manual and automatic, match conditions, counterpart items, standard models. In plain language.
What is IP transit?
IP transit explained: how your network reaches the rest of the internet through a transit provider. On autonomous systems, BGP, the full table and multihoming.
The cash flow statement in Odoo Accounting
Cash flow in Odoo explained: opening balance, operating, investing, account tags. In plain language.
Answer not found?
Ask an engineer directly — we usually respond within one business day.