Cloud

"Cloud is the new OS"

Cloud is just a fancy name for systems (in this modern world, they are most likely distributed systems), with the hardware procurement and maintenance and many other things abstracted away from software developers.

Cloud can be public, meaning Amazon or Microsoft or Google is running your infrastructure. Or private, meaning you still fully own your data center but resource request and allocation is done through software UI. Or hybrid, taking the best parts of public and private cloud.

Based on the abstraction level, cloud offerings can be infrastructure as a service (IaaS), platform as a service (PaaS), and software as a service (SaaS).

Computation evolution: bare metal -> virtualized -> containerized -> cloud -> serverless.

Cloud Infrastructure

3 big players (60%+ combined market share):

• Amazon AWS
• Microsoft Azure
• Google Cloud Platform

Other players (less than 10% of market share each):

• Outside of China:
  • Oracle
  • IBM / Redhat
• China: (the top 4 has >80% marketshare in China)
  • Alibaba Cloud
  • Huawei Cloud
  • Tencent Cloud
  • Baidu AI Cloud

Specialized:

• Okta: identity.
• DigitalOcean: compute.
• Backblaze: storage / backup.
• Snowflake: data warehouse.
• Hashicorp: infrastructure-as-code (IaC), enable IT collaboration and automation.
• CloudFlare: CDN, security.
• ServiceNow: task management.
• PagerDuty: Incident Response.
• Workday
• Salesforce

3 Key Categories

Each public cloud offers dozens of services. To make it easier to understand, think of these 3 key categories:

• Compute: VM, containers, serverless functions, etc.
• Storage: databases, datawarehouses, object stores, etc.
• Networking: DNS, VPC, load balancing, etc.

Virtualization

Virtualization is the key enabler of the Cloud, so that the resources can be split and re-packaged to sell: virtual machine (hypervisor), virtualized (software-defined) storage and network.

Hyper-converged infrastructure (HCI): software-defined infrastructure, virtualizes all of the elements of conventional "hardware-defined" systems. HCI is mostly a marketing term.

Read more about Virtualization.

How to provision cloud resources

• CLI (e.g. aws, gcloud).
• Web UI (e.g. AWS Management Console, Google Cloud Concole).
• REST API.
• Infrastructure-as-code tools
  • AWS has CloudFormation.
  • Terraform can work on multiple clouds; the license change may affect its adoption.
  • GCP has Infrastructure Manager which utilizes Terraform.

Resource Hierarchy Comparison

• GCP: Organization => Folder => nested Folder => Project => Resource
• AWS: Organization => OrganizationUnit => nested OU => Account => Resource
• Azure: Active Directory Tenant (Root Management Group) => Management Group => nested Management Group => Subscription => ResourceGroup => Resource

Events / Conferences

Follow the latest developments in the Cloud world:

• AWS re:Invent: usuall at the end of the year, in Las Vegas.
• Google Cloud Next: usually at the end of summer; used to be in San Francisco, moved to Las Vegas in 2024.

Disks

• AWS: Amazon Elastic Block Store (EBS).
• GCP: Persistent Disk (PD). AWS uses the term "volume" while GCP uses "disk."
• Azure: Managed Disks. Azure uses the term "managed disks" while GCP uses "persistent disks."

Comparison Charts

https://cloud.google.com/docs/get-started/aws-azure-gcp-service-comparison

How public clouds tak to each other?

In the marketing world, AWS and Google Cloud Platform (GCP) are fierce rivals. But in the physical world—specifically inside the humming, chilled hallways of Equinix IBX (International Business Exchange) data centers—these giants are neighbors who share the same "sidewalks."

For a modern enterprise, "The Cloud" is rarely just one company. Most large-scale architectures are multicloud, meaning they might run AI models on Google Vertex AI while storing the primary database on AWS RDS. For this to work without a 50-millisecond lag, the clouds cannot talk over the public internet; they must physically "shake hands."

Here is how the cloud-to-cloud interconnection works, from the physical fiber to the software-defined future.

The Neutral Ground: Switzerland in a Server Rack

If AWS is a kingdom and GCP is an empire, Equinix is the neutral "International Zone."

Equinix IBX centers are colocation facilities. They don't sell cloud services; they sell space, power, and, most importantly, interconnectivity. Inside a massive Equinix facility (like Ashburn, VA or Frankfurt, Germany), AWS has cages full of routers, and Google has cages just a few hundred feet away.

The "handshake" happens in the Meet-Me Room (MMR). This is a central room where thousands of fiber optic cables terminate. When a customer wants AWS and GCP to talk:

1. A technician literally runs a yellow fiber-optic cable (a Cross-Connect) from the AWS router rack to a patch panel.
2. Another cable is run from the GCP rack to the same panel.
3. They are plugged together.

At this moment, the "Cloud" is no longer a concept; it is a physical light signal traveling across a piece of glass at 186,000 miles per second.

The Traditional Way: The "DIY" Interconnect

Previously, if you wanted AWS and Google to talk, you had to act as the "middleman." This usually involved:

• AWS Direct Connect (DX): Your dedicated private pipe into Amazon’s network.
• Google Cloud Interconnect (GCI): Your dedicated pipe into Google’s network.

You would hire a third party (like Equinix, Megaport, or PacketFabric) to place a "virtual router" between the two. This router would handle the BGP (Border Gateway Protocol) peering—the "language" that tells AWS, "Hey, if you have traffic for Google's IP addresses, send it through this cable."

The 2026 Shift: AWS Interconnect (Multicloud)

We are currently seeing a paradigm shift where the clouds are finally making it "point-and-click." The AWS Interconnect (Multicloud) feature (introduced as a standard by 2026) represents the "Software-Defined" era of peering.

Instead of calling Equinix or managing your own BGP routers, AWS has pre-provisioned massive "fat pipes" to Google and Azure.

• How it works: Right from the AWS Management Console, you can provision a "Cloud-to-Cloud Link."
• Under the hood: AWS uses its existing physical presence in neutral data centers to logically partition a slice of bandwidth for your specific VPC.

This effectively turns the "Great Wall" between clouds into a "Door." You no longer need to be a networking expert to link an AWS Lambda function to a Google BigQuery dataset.

Why This Matters: The Three "S" Pillars

Why go through all this trouble instead of just using the internet?

A. The Egress Tax (Savings)

Moving data out of a cloud ("Egress") is notoriously expensive. However, when clouds talk over a direct interconnect in a facility like Equinix, they often offer reduced egress rates. It turns the "Hotel California" model (you can check out, but you can never leave) into a more fluid "City Transit" model.

B. Latency (Speed)

Traffic over the public internet is "jittery." It might hop through five different ISPs before reaching its destination. A direct cloud-to-cloud link reduces latency from 50ms+ to as little as 1ms to 2ms. For high-frequency trading or real-time AI inference, this is the difference between a functional product and a broken one.

C. Security (Silence)

When you use a direct interconnect, your data never touches the public internet. It stays entirely within private fiber. This eliminates the risk of DDoS attacks or packet sniffing at the ISP level, making it the only viable option for government, healthcare, and finance sectors.

The Future: The Multicloud Mesh

As AWS Interconnect and similar tools from Google and Azure become standard, the "Cloud" is evolving into a distributed mesh.

The physical reality remains: your data is still sitting on a Linux server (likely an ELF binary running on a kernel we discussed earlier) inside an Equinix building. But the logical reality is shifting toward a world where the brand name of the cloud provider matters less than the speed of the handshake between them.