Insights
Plain-language engineering notes on private cloud, self-hosted AI, security at the edge, networks and voice.
Monitoring that tells you something: status pages and alerts worth having
Most monitoring produces either silence or noise. What a useful status page measures, why percentiles beat averages, and how to build alerting that someone actually acts on.
SD-WAN for multi-site Australian businesses: resilience over NBN, fibre, Starlink and 4G
What SD-WAN is, how it keeps sites connected across NBN, fibre, Starlink and 4G, and the situations where a simple VPN is the better answer.
Sizing GPU servers for self-hosted AI: models, users, retrieval and headroom
GPU memory, more than raw compute, decides what a self-hosted AI server can do. How to size for the models you want, the users you actually have, document retrieval and room to grow.
Hosted PBX done properly: what actually changes for a 20 to 200 seat business
Multi-tenant hosted PBX with a real self-service portal, WebRTC softphones, queues and recording, and a sane migration path off the old on-prem system.
Blocking at the edge: what IPS/IDS with ML threat scoring does and does not do
A firewall opens ports. An IPS inspects what comes through them. What edge blocking with ML threat scoring stops, what it misses, and why logs matter.
Applying the Essential Eight on infrastructure you own
The ASD's Essential Eight was written with Windows networks in mind. When you also own the hypervisors, storage and management plane, here is how its principles extend to the layers underneath.
Self-hosted enterprise AI: GPT-class tools without sending documents offshore
How to give staff capable AI tools while documents stay inside your boundary: multi-model routing, document-native workflows, governance and cost visibility.
Backups that survive ransomware: immutability, isolation and restores you have actually tested
Ransomware operators go looking for the backups first. What it takes for a backup to survive: copies that cannot be changed, credentials the attacker cannot reach, and restores that have been rehearsed.
Hyperconverged infrastructure explained: why the default install disappoints
Compute and storage in one chassis is simple to describe and easy to get wrong. What HCI is, how distributed storage works, and the tuning that matters.
Leaving VMware: a practical path to open hyperconverged infrastructure
Many organisations are reviewing VMware at renewal time. How to plan a move to open-source hyperconverged infrastructure without betting everything on one weekend.
Private cloud vs public cloud: what actually matters for Australian organisations
Sovereignty, predictable cost and latency decide the private-versus-public question workload by workload. Here is how to make the call honestly.