[All posts](https://zsoftly.ca/blog)
Blog

# Engineering

Architecture decisions, performance, and how we build ZCP.

Engineering

September 23, 2026

### [Why Use Private Desktops?](https://zsoftly.ca/blog/why-we-built-private-employee-desktops)

A business case for private employee desktops: controlled backend access from everyday employee devices.

Engineering

September 19, 2026

### [Even On-Prem, Servers Should Be Cattle, Not Pets](https://zsoftly.ca/blog/even-on-prem-servers-should-be-cattle-not-pets)

How we rebuilt our authoritative DNS design around independently operable roles, a reversible cutover, and observability that describes the system we now run.

Engineering

September 16, 2026

### [We Moved Our Own DNS onto Our Own Platform](https://zsoftly.ca/blog/migrating-our-dns-to-our-own-platform)

How we registered .ca nameserver host records, recovered from an unrequested delegation change, and moved zsoftly.ca onto the nameservers we sell to customers.

Engineering

August 22, 2026

### [Local vs Replicated Block Storage in YUL-1: Performance, Price, and Recovery](https://zsoftly.ca/blog/local-vs-distributed-yul-root-storage-results)

A YUL-1 comparison of local NVMe, local SATA SSD, and shared replicated block storage. See the performance results, pricing model, and recovery tradeoffs before choosing a storage tier.

Engineering

August 20, 2026

### [What Nobody Tells You About Migrating a VPN Server From AWS to Private Cloud](https://zsoftly.ca/blog/migrating-a-vpn-server-from-aws-to-private-cloud)

A working public-cloud deployment does not always survive the move to private infrastructure. Here is the failure pattern we hit moving a VPN control plane off AWS, and the lessons that apply to any public-to-private cloud migration.

Engineering

August 16, 2026

### [What a Private AI Workload Taught Us About Compute, Cost, and Control](https://zsoftly.ca/blog/what-we-learned-running-ollama-on-zcp-intel-yul)

We tested Ollama and Open WebUI on a 64 GB Intel VM in ZCP YUL-1. Here is what decision-makers should know about private AI, model size, cost, speed, and deployment fit.