Client-Server Model

Deployment Models

You can think as deployment models as:

• How can you integrate the cloud resources with the infrastructure of the client.

• Cloud-based
  • Run all parts of the app in the cloud.
  • Migrate existing app entire to the cloud.
  • Design and build new apps in the cloud.
  • Use Cases:
    • The clients don’t need full control over the infrastructure and data.
    • The workloads can run entirely in the cloud.

• On-premises (Private cloud)
  • The first approach is the clients provide the infrastructure and AWS provide the software layer. Case the clients want to run optimized apps and services instead of create legacy app.
  • The second approach is AWS literally provide real infrastructure to your site, but AWS still manages and updates it.
    • See AWS Outposts Family.
  • Think as the clients have their own running infrastructure and they want to increase their resources and software services, but they want to keep the physical servers in site.
  • The client BUY and MANAGE hardware. Handle power, cooling, networking, backups, and disaster recovery.
  • Use Cases:
    • Highly sensitive systems (e.g., certain government, finance, healthcare environments)
    • Legacy app tightly coupled to specific hardware.

• Hybrid (AWS + On-Premise)
  • Think as the clients want to run some workloads in their own infrastructure and others in AWS cloud servers., and often connect them so they work together.
  • Use Cases:
    • Keep sensitive systems on-premise, and non-sensitive in the AWS servers.
    • Gradual cloud migration strategies.
    • Improved disaster recovery and backup options.
    • Seasonal workloads needing burst capacity.
    • Keeping regulated data on-prem while running analytics in AWS.

Benefits of using compute in the cloud

• Buy, keep, and maintain physical servers can be expense. Using cloud computing resources you pay what you consume.

• Flexibility:
  • You don’t care about keep the infrastructure updated.
  • You don’t need to add manually more resources if the apps need more.

• Deploy apps to the customers quickly in any place with low latency.

• The points before lead to scale your business.

AWS services to provide cloud computing

Amazon Elastic Compute Cloud (Amazon EC2)

• It’s a virtual server with computing resources associated to it.

Types of Amazon EC2 instances

• General purpose instances
  • Good balance between compute, memory, storage, and network resources.
  • Use cases:
    • App servers
    • Gaming servers
    • Backend servers for enterprise apps
    • Small and medium databases.

• Compute optimized instances
  • Idead for high demanding CPUs apps.
  • Use cases:
    • High-performance web servers.
    • Compute intensive app servers.
    • Dedicated gaming servers.
    • Batch processing workloads.

• Memory optimized instances
  • High RAM capacity instances.
  • Use cases:
    • Process large datasets.
    • Real time processing of a large amount of unstructured data.

• Storage optimized instances
  • Designed for workloads that demands high, sequential read and write access to large datasets on local storage.
  • Use cases:
    • Data warehousing apps.
    • High frequency online transaction processing (OLTP) systems.

• Accelerated computing instances
  • Use hardware accelerators, coprocessors, or GPUs to perform some functions more efficiently than is possible in software running on CPUs.
  • Use cases:
    • Graphic processing.
    • Floating-point calculations.
    • Data pattern matching.

Pricing

Main principle → PAY-AS-YOU-GO

• On-demand:
  • The instance don’t stop until you stop it.
  • Pay only the compute time you use.
  • Ideal for short terms.

• Reserved Instances:
  • You purchase a/a group of instances for 1-year or 3-year term.
  • There are 2 types:
    • Standard Reserved Instances: Good choice if you know:
      • Instance type.
      • Instance size.
      • The region to deploy the app.
    • Convertible Reserved Instances:
      • Good fit if you need to run the app in different Availability Zones or different instances types.

• Spot Instances:
  • Ideal for workloads with flexible start and end times.
  • Used typically for background process without affecting the overall operations of your business.
  • Offer 90% discount of On-demand price.

• Dedicated Hosts:
  • Most expensive option.
  • Physical servers with Amazon EC2 instances that are fully dedicated to your use.

• EC2 Instance Saving Plan:
  • Provide a discount when you commit to make a hourly spend to an instance family and Region for a 1-year or 3-year term.

Scalability (Amazon EC2 Auto Scaling)

• Amazon EC2 Auto Scaling provide scalability for Amazon EC2 instances. Think as increase the number of instances while you app need it.

• You can use 2 approaches:
  • Dynamic scaling: Responds to changing demand.
  • Predictive scaling: Automatically schedules the right number of Amazon EC2 instances based on predicted demand.

• You have to set the Auto Scaling Group. You set the following set up:
  • Minimum: Number of instances that need your app to run.
  • Desired: Number of instances that you want to use in your app.
  • Maximum: Number of instances that you reserve in case the app need more resources.

  

Elastic Load Balancing

• Service that automatically distribute the incoming traffic across multiple resources such as Amazon EC2.

• Amazon EC2 assigns instances - Elastic Load Balancing distributes to the assigned instances.

Messaging and Queuing

The apps can have two main architecture design:

• Tightly coupled architecture.
  • When component A fail, component B also fail.
  • Think as all the app in a single project.
  • You don’t need to send messages to the app components through the infrastructure.

• Loosely coupled architecture.
  • Think as microservices where each microservice has its own function.
  • So, if component A fail, component B don’t fail.
  • You need a message service to communicate the components.

There are 2 AWS services to send messages and handling queues.

• Amazon SNS (Simple Queue Service)
  • It’s a publisher/subscriber service. More native way of send messages one side publish a message and the other side listen the message.
  • You can loss messages, because there isn’t a buffer between.
  • Use cases:
    • Web servers
    • AWS Lambda functions.

• Amazon SQS (Simple Notification Service)
  • It’s a message queuing service.
  • You don’t loss messages because they wait in a buffer or queue.

Serverless and Containers Solutions

Serverless → Don’t care about managing the server

• With Amazon EC2 you need:
  • Virtual server + Code + Manage the instance.
  • No native serverless approach.

• With serverless approach you need only the Code part or only need to focus on the product not the infrastructure.

AWS Lambda

• It’s a service that lets you run code without worrying about managing the infrastructure. (serverless approach)

• How AWS Lambda work:

Containers

• Amazon Elastic Container Service (Amazon ECS)
  • Management system that allow build, run and scale containerized apps on AWS.

• Amazon Elastic Kubernetes Service (Amazon EKS)
  • Management system to orchestrate containers on AWS.

• AWS Fargate
  • It’s a serverless compute engine for containers on Amazon ECS or Amazon EKS to run containerized apps without worrying about the server management or instance allocation.

Summary

• Your business can choose from 3 deployment models options. In any case, you will end up using AWS services.

• Depends of you business needs or the architecture of your apps you can use the CORE and add some necessary services.

• If your business needs or app scale maybe you need to add more services.

• If your business doesn’t need to worry about managing infrastructure, you can opt for a serverless approach.