ANSAI TECHNOLOGIES × GDG TUKGoogle Developer Group · Technical University of Kenya
The Cloud, From First Principles
Session 1 · Cloud Foundations — what the cloud actually is at the level of hardware and software, how one is built, and how to begin on Google Cloud without billing risk.
AI has collapsed much of building. Generating an app is now cheap — almost free. What did not become cheap is judgment: knowing what's underneath, what's true, and what will still stand in five years
That's why fundamentals matter more now, not less. Engineering is a craft — a way of thinking clearly about real things, and taking responsibility for what you make
And data sovereignty: our data lives on someone's land, under someone's law. If Kenya's digital future is going to be built by Kenyans, it needs engineers who understand foundations — not just tenants of other people's platforms
Tools change. Thinking compounds. That's the whole philosophy — tonight is an exercise in it.
Ansai Technologies × GDG TUK3
Tonight's plan
By the end of tonight, you will be able to…
Define cloud computing precisely — resources, mechanism, and billing model
Specify how a small cloud is built: the components, and the software that runs it
Provision a Google Cloud account and configure billing alerts correctly
Reason about data residency — where data physically sits, and under whose jurisdiction
And don't worry: no servers will be harmed in this session. 😄
Ansai Technologies × GDG TUK4
First, some good news
You already use the cloud. Every day.
Gmail — your mail is stored on Google's servers and synced to your phone; the phone is a window, not the warehouse
Google Drive & Photos — lose the phone and the photos survive, because the authoritative copy lives in a data centre
WhatsApp, M-Pesa, Netflix — every one is software executing on remote infrastructure you never see
So the cloud is not new to you. Tonight we examine the machinery behind it.
Ansai Technologies × GDG TUK5
Follow one request, precisely
What happens when you open a website
DNS resolution: your device asks a resolver; an A/AAAA record returns the server's IP address
Connection: TCP opens, TLS encrypts; packets are routed across your ISP and submarine fibre
Processing: the server executes code (compute) against stored state (storage) and returns a response — the full round trip is often under 300 ms
Ansai Technologies × GDG TUK6
So where does the cloud live?
In buildings engineered for exactly this
A data centre: thousands of servers (rack-mounted computers), redundant power (grid, UPS, generators), industrial cooling, and physical security — on real land, under a real jurisdiction.
Cloud computing is the delivery of computing services — servers, storage, databases, networking, software — over the internet from such facilities, metered: you pay for what you use.
Ansai Technologies × GDG TUK7
The components
Four ingredients. That's the whole recipe.
🧠 Compute
Processors and memory — the execution of instructions. Provisioned and billed as vCPUs and GiB of RAM.
💾 Storage
Persistent media — block, file and object storage on NVMe, SSD and HDD. The state that survives power loss.
🛣️ Networking
Links, switching and routing; IP addressing and DNS. Its measurable properties are bandwidth and latency.
⚡ Power & Cooling
The physical plant — grid feed, UPS, generators, and heat removal. Data centres report it as PUE efficiency.
Ansai Technologies × GDG TUK8
Mechanism one
One computer, split into many
A hypervisor partitions one physical host's CPU, memory, storage and network I/O among several guest machines
Each guest runs its own operating system kernel on allocated, enforced resources
Isolation is enforced by the hypervisor: guests share hardware — they never share state
Containers share the host kernel instead; isolation comes from namespaces and cgroups — lighter, with a different trust boundary
The intuition: a block of flats — one structure, self-contained units. The mechanism, though, is exact partitioning, enforced in software.
Physical host — partitioned live
Guest VM 1
Own kernel · 2 vCPU · 4 GiB RAM
Guest VM 2
Own kernel · 2 vCPU · 4 GiB RAM
Guest VM 3
Own kernel · 4 vCPU · 8 GiB RAM
Guest VM 4
Own kernel · 1 vCPU · 2 GiB RAM
HYPERVISOR — Type 1 · KVM / ESXi / Hyper-V
HARDWARE — CPU · RAM · NIC · DISK
Ansai Technologies × GDG TUK9
Mechanism two
Provisioning as an API call
Every resource — a VM, a disk, a network — is created, configured and destroyed through an API
Infrastructure as Code (Terraform, the gcloud CLI) declares the desired state; the platform reconciles reality to match it
The result: a server in seconds, elasticity on demand, and usage metered — the basis of pay-as-you-go
The web console you will use tonight is one client of that API. Anything the console does, code can do
Ansai Technologies × GDG TUK10
The definition, assembled
Computers + virtualization + automation + a network = THE CLOUD
Formally (NIST SP 800-145): pooled resources, broad network access, on-demand self-service, rapid elasticity, measured service. Google has more computers than you do. If the formula holds… could you build one?
Ansai Technologies × GDG TUK11
A true story
I built one — from an old laptop
An old Compaq laptop, retired from service, now runs as a home cloud node: DC-01, on Debian Linux
AdGuard Home — network-wide DNS filtering · WireGuard — an encrypted VPN tunnel back home
Syncthing — continuous file synchronization (a private Drive) · Uptime Kuma — service health monitoring
Cost: electricity. Operator and landlord: me
Ansai Technologies × GDG TUK12
Your turn
Build your own cloud, in five steps
1. Take any old computer — it already contains compute, storage, networking and power
2. Install Linux (Debian or Ubuntu Server) — the operating system of nearly all cloud infrastructure
3. Add services, all free and open source: file sync, VPN, DNS filtering, media, backups
4. Address it: a static IP and a hostname on your network — your own small DNS entry
5. Go further when ready: a hypervisor (Proxmox / KVM) turns your one machine into a host of many
Ansai Technologies × GDG TUK13
Why location is an engineering decision
Data sovereignty: whose soil is your data on?
Data residency is physical: your bytes sit on disks in a specific building. Jurisdiction follows: Kenyan data answers to the Data Protection Act (2019); data abroad answers elsewhere
Renting offshore means rent in foreign currency, forever — and a control plane in someone else's hands
Owning the infrastructure means control and duty: redundancy, patching, backups and uptime become yours
Competent engineers decide per workload — residency, latency, cost and law weighed together
Ansai Technologies × GDG TUK14
And now — the cherry on top 🍒
Google Cloud: the same picture, at world scale
The ingredient
At your desk
At Google
Compute
A laptop CPU
Compute Engine — VMs (IaaS)
Storage
A hard disk
Persistent Disk (block) · Cloud Storage (object)
Networking
A Wi-Fi router
VPC — software-defined networking
The facility
Your room
A region — independent geographic area; zones are isolated failure domains within it
The same infrastructure that serves Search, Gmail and YouTube, opened to builders. Our nearest region is africa-south1 (Johannesburg) — light has a speed limit, so distance is latency; and remember the philosophy: understand what the console is, don't just memorize it.
Ansai Technologies × GDG TUK15
Let's get hands-on 🛠️
The console, in three ideas
📁 Projects
The unit of isolation and billing. Every resource belongs to exactly one project; quotas, IAM and billing attach at this level.
🔑 IAM
Policies bind principals (users, groups, service accounts) to roles (permission sets). The operating rule is least privilege: grant the minimum a task requires.
And the console itself is a web client over the same APIs you just learned. Let's open it: console.cloud.google.com
Ansai Technologies × GDG TUK16
Billing safety — the guardrails, stated exactly
The free trial, with guardrails
New accounts receive a free trial: $300 credit, 90 days. A card is required for identity verification; no charge occurs during the trial, and resources stop when it ends unless you upgrade
Configure a budget with alerts at 50%, 90% and 100% of actual (or forecasted) spend — before creating any resource
Precision matters: a budget alert is a notification, not a cap. Enforcement is programmatic — an alert can trigger automation that disables billing. An alert alone only warns
The operating habit that prevents every surprise bill: delete resources you are not using
Ansai Technologies × GDG TUK17
Before Session 2
Your homework 📝
Provision: complete your Google Cloud account registration
Protect: configure your budget alerts and verify they are active
Analyse: take any old computer at home and inventory it as infrastructure — CPU and RAM (compute), disk capacity and type (storage), network interface (networking), power supply (power). What service could it run?
Session 2 assumes verified consoles — and, I hope, a few of those inventories.
Ansai Technologies × GDG TUK18
One more door, for the hungry ones
The Ansai Apprenticeship
Everything tonight rests on one belief: engineering is a craft of thinking. At Ansai we train that craft properly — real projects, a hard problem every week, a mentor reviewing your work, until you can take a problem from first sketch to shipped system.
A new cohort is forming. If tonight lit something in you, register your interest — it takes two minutes:
apply.ansaitechnologies.co.ke
Ansai Technologies × GDG TUK19
Session 1 · Wrap-up
It's computers all the way down. Now you know what's under the button.
Asante sana! Questions are welcome — and as you leave, tell me: what will your first cloud run?
