Cloud Computing

Service delivery model over the internet (cloud). This includes but is not limited to

• compute power meaning servers such as Windows, Linux, hosting environments, etc. storage like files and/or databases
• networking in azure but also outside when connecting to your company network - analytics services for visualization and telemetry data

Key concepts

• scalability is the ability to scale, so allocate and deallocate resources at any time
• elasticity is the ability to scale dynamically
• agility is the ability to react fast (scale quickly)
• fault tolerance is the ability to maintain system uptime while physical and service component failures happen
• disaster recovery is the process and design principle which allows a system to recovers from natural or human-induced disasters
• high availability is the agreed level of operational uptime for the system. It is a simple calculation of system uptime versus the whole lifetime of the system.

availability = uptime/(uptime + downtime)

Economies of Scale

The principle of economies of scale states that as the companies grow they become more effective at managing shared operations. Be that HR and hiring, taxes, accounting, internal operations, marketing, big purchases via contracts meaning better discounts, etc.

Because of those, companies can save/earn more which in return allows for a reduction in the cost of their services to their customers. This is the so-called ‘price per unit.

It’s not possible to go to 0 because in the end, some underlying infrastructure needs to run to provide the services. But the larger the scale the more benefits can be passed to customers.

In fact, at the current scale, Microsoft can already offer multiple services for free due to how small a fraction of the cost it is for them.

CapEx vs OpEx

Differences between Capital Expenditure and Operational Expenditure

What is a consumption-based model?

The consumption-based model is a pricing model used in the cloud so that customers are only charged based on their resource usage.

This model is characterized by

• No associated upfront cost
• No wasted resources as such no charges are incurred for unused resources*. Unused in this case is different per service. For instance, blob storage that stores any data is considered to be used, as it consumes the storage space. Virtual Machines that are running consume CPU, memory and other resources even if there isn’t any traffic. Hence they are considered to be used and will incur charges.
• Pay for what you need
• Stop paying when you don’t

Consumption is the virtual metric used to calculate how much each resource (service) in Azure was used. Each service has many smaller metrics that track its consumption to offer best possible pricing model. Those metrics are tracked on very granular level.

Service Models responsibilities

As a service means which party will manage the particular layer and all the layers below.

• Software layer consists the application (application code and set) & the application data
• Platform layer means all the supporting software and the operating system required to host the application
• Infrastructure layer consists hardware the infrastructure and virtualization required to host the platform

Responsibility Matrix

As such following table represents the responsibilities

Cloud Deployment Model

Cloud Deployment Model is simple a separation that describes where are the company resources deployed. Whenever this is in a public cloud provider environment or private data center.

The below table presents high-level deployment model separation

Public Cloud

Private Cloud

Hybrid Cloud

Data Center

• Physical facility
• Hosting for a group of networked servers
• Own power, cooling & networking infrastructure

Region

• Geographical area on the planet
• One but usually more datacenters connected with low-latency network (<2 milliseconds)
• Location for your services
• Some services are available only in certain regions
• Some services are global services, as such are not assigned/deployed in specific region
• Globally available with 50+ regions
• Special government regions (US DoD Central, US Gov Virginia, etc.)
• Special partnered regions (China East, China North)

Availability Zone

• Regional feature
• Grouping of physically separate facilities
• Designed to protect from data center failures
• If zone goes down others continue working
• Two service categories Zonal services (Virtual Machines, Disks, etc.) Zone-redundant services (SQL, Storage, etc.)
• Not all regions are supported -Supported region has three or more zones
• A zone is one or more data centers

Region Pair

• Each region is paired with another region making it a region pair
• Region pairs are static and cannot be chosen
• Each pair resides within the same geography* Exception is Brazil South
• Physical isolation with at least 300 miles distance (when possible)
• Some services have platform-provided replication
• Planned updates across the pairs
• Data residency maintained for disaster recovery

Geographies

• Discrete market
• Typically contains two or more regions
• Ensures data residency, sovereignty, resiliency, and compliance requirements are met
• Fault tolerant to protect from region-wide failures
• Broken up into areas Americas, Europe, Asia Pacific, Middle East and Africa
• Each region belongs only to one Geography

Azure Resource

• Object used to manage services in Azure
• Represents service lifecycle
• Saved as JSON definition

Resource Groups

• Grouping of resources
• Holds logically related resources
• Typically organizing by : Type, Lifecycle (app, environment), Department, Billing, Location or combination of those

Resource Manager

• Management Layer for all resources and resource groups
• Unified language - Controls access and resources

Additional Info

• Each resource must be in one, and only one resource group
• Resource groups have their own location assigned
• Resources in the resource groups can reside in a different locations
• Resources can be moved between the resource groups
• Resource groups can’t be nested
• Organize based on your organization needs but consider: Billing Security and access management Application Lifecycle

Virtualization

• Emulation of physical machines
• Different virtual hardware configuration per machine/app
• Different operating systems per machine/app
• Total separation of environments: file systems, services, ports, middleware, configuration

Virtual Machines

• Infrastructure as a Service (IaaS)
• Total control over the operating system and the software
• Supports marketplace and custom images
• Best suited for: Custom software requiring custom system configuration, Lift-and-shift scenarios
• Can run any application/scenario: web apps & web services, databases, desktop applications, jumpboxes, gateways, etc.

Virtual Machine Scale Sets

• Infrastructure as a Service (IaaS)
• Set of identical virtual machines
• Built-in auto scaling features
• Designed for manual and auto-scaled workloads like web services,* batch processing, etc.

Containers

• Use host’s operating system
• Emulate operating system (VMs emulate hardware)
• Lightweight (no O/S): Development Effort Maintenance Compute & storage requirements
• Respond quicker to demand changes
• Designed for almost any scenario

Azure Container Instances

• Simplest and fastest way to run a container in Azure
• Platform as a Service
• Serverless Containers
• Designed for: Small and simple web apps/services Background jobs Scheduled scripts

Azure Kubernetes Service (AKS)

• Open-source container orchestration platform
• Platform as a Service
• Highly scalable and customizable
• Designed for high scale container deployments (anything really!)

App Service

• Designed as enterprise grade web application service
• Platform as a Service
• Supports multiple programming languages and containers

Azure Functions (Function Apps)

• Platform as a Service
• Serverless
• Two hosting/pricing models: Consumption-based plan Dedicated plan
• Designed for micro/nano-services

Summary

• Virtual Machines (IaaS) - Custom software, custom requirements, very specialized, high degree of control
• VM Scale Sets (IaaS) - Auto-scaled workloads for VMs
• Container Instances (PaaS) - Simple container hosting, easy to start
• Kubernetes Service (PaaS) - Highly scalable and customizable * container hosting platform
• App Services (PaaS) - Web applications, a lot of enterprise web * hosting features, easy to start
• Functions (PaaS) (Function as a Service) (Serverless) - micro/nano-services, excellent consumption-based pricing, easy to start

Azure Networking

• Connect cloud and on-premises
• On-premise networking functionality

Azure Virtual Network

• Logically isolated networking components
• Segmented into one or more subnets
• Subnets are discrete sections
• Enable communication of resources with each-other, internet and on-premises
• Scoped to a single region
• VNet peering allow cross region communication
• Isolation, Segmentation, Communication, Filtering, Routing

Azure Load Balancer

• Even traffic distribution
• Supports both inbound and outbound scenarios
• High-availability scenarios
• Both TCP (transmission control protocol) and UDP (user datagram protocol) applications
• Internal and External traffic
• Port Forwarding
• High scale with up to millions of flows

VPN Gateway

• Specific type of virtual network gateway for on-premises to azure traffic over the public internet

Application Gateway

• Web traffic load balancer
• Web application firewall
• Redirection
• Session affinity
• URL Routing
• SSL termination

Content Delivery Network

• Define content
• Minimize latency
• POP (points of presence) with many locations

Data Types

• Structured - Data that can be represented using tables with very strict schema. Each row must follow defined schema. Some tables have defined relationships between them. Typically used in relational databases.
• Semi-structured - Data that can be represented using tables but without strict defined schema. Rows must only have unique key identifier.
• Unstructured - Any files in any format. Like binary files, application files, images, movies, etc.

Storage Account

• Group of services which include: blob storage, queue storage, table storage, and file storage
• Used to store: files, messages, and semi-structured data
• Highly scalable (up to petabytes of data)
• Highly durable (99.999999999% - 11 nines, up to 16 nines)
• Cheapest per GB storage

Blob Storage

• BLOB – binary large object – file
• Designed for storage of files of any kind
• Three storage tiers: Hot – frequently accessed data Cool – infrequently accessed data (lower availability, high durability) Archive – rarely (if-ever) accessed data

Queue Storage

• Storage for small pieces of data (messages)
• Designed for scalable asynchronous processing

Table Storage

• Storage for semi-structured data (NoSQL): No need for foreign joins, foreign keys, relationships or strict schema Designed for fast access
• Many programming interfaces and SDKs

File Storage

• Storage for files accessed via shared drive protocols
• Designed to extend on-premise file shares or implement lift-and-shift scenarios

Disk Storage

• Disk emulation in the cloud
• Persistent storage for Virtual Machines
• Different: sizes, types (SSD, HDD) performance tiers
• Disk can be unmanaged or managed