Optimizing Compute in AWS

AWS Cloud Technology and Services

Rahulraj Singh

Technical Product Manager

What is compute in AWS?

AWS compute = virtual processing power

Three images highlighting cloud servers, physical machines, and power backup.

What is Amazon Elastic Compute Cloud (EC2)?

Provides resizable compute capacity in the cloud
Each individual EC2 machine is referred to as an instance

Key characteristics:

Ability to scale up or down based on demand
Availability of a varied range of instance types for specialized use cases

A branched chart showing compute connection to storage, databases, and networks.

EC2 instance type categories

AWS offers six categories of EC2 instances for specialized workloads

General purpose
Compute optimized
Memory optimized
Storage optimized
Accelerated computing
High Performance Computing (HPC) optimized

Icons showing grouped EC2 instances and a singular EC2 instance.

General purpose and storage optimized instances

General purpose instances

Balance of compute, memory, and networking resources
Use cases:
- Hosting dynamic websites
- Maintaining code repositories

EC2 instance icon showing a general purpose instance.

Storage optimized instances

High, sequential read and write access to large datasets
Use cases:
- Data warehousing
- Refactoring large relational databases

EC2 instance icon showing a storage optimized instance.

Compute and memory optimized instances

Compute optimized instances

Compute-intensive and high-performance workloads
Use cases:
- Scientific simulations
- Financial modeling

EC2 instance icon showing a compute optimized instance.

Memory optimized instances

Memory-intensive workloads
Use cases:
- Real-time stream data analytics
- Generating close captions

EC2 instance icon showing a memory optimized instance.

Specialized compute instances

Accelerated computing instances

Contain specialized hardware accelerators, like GPUs or FPGAs
Use cases:
- Deep learning
- Rendering gaming graphics

EC2 instance icon showing an accelerated computing instance.

HPC optimized instances

Best price performance for running high performance workloads at scale
Use cases:
- Weather forecasting
- Crash simulations

EC2 instance icon showing a high performance compute optimized instance.

Load balancing in AWS

Load balancing ensures even distribution of incoming traffic among multiple EC2 instances, preventing overload on a single server

Ensures high availability
Provides horizontal scaling

Types of load balancers in AWS

Four icons showing classic, network, application, and gateway load balancers.

How does load balancing work?

Users send requests

An architecture diagram highlighting user requests sent to a load balancer.

How does load balancing work?

Users send requests
Requests hit the load balancer

An architecture diagram highlighting user requests sent to a load balancer.

How does load balancing work?

Users send requests
Requests hit the load balancer
Primary target group is instantiated by the application load balancer first

An architecture diagram highlighting a user request move from a load balance to primary EC2 instances.

How does load balancing work?

Users send requests
Requests hit the load balancer
Primary target group is instantiated by the application load balancer first
If demand increases, the load balancer activates the secondary target group and distributes the load across all instances

An architecture diagram highlighting a user request move from a load balance to primary and secondary EC2 instances.

What is compute elasticity?

Elasticity ensures your system can scale up or down based on demand, providing flexibility in resource allocation

EC2 instances achieve elasticity through EC2 Auto Scaling

What is EC2 Auto Scaling?

Automatically adjust the number of active instances based on usage and requirement
Optimize costs
Prevent over-provisioning

How does auto-scaling work?

Users send requests

An architecture diagram showing a user request generation.

How does auto-scaling work?

Users send requests
The requests are routed to EC2 Auto Scaling service

An architecture diagram showing a user request hitting an EC2 Auto Scaling service.

How does auto-scaling work?

Users send requests
The requests are routed to EC2 Auto Scaling service
The service then routes requests to the active EC2 instances

An architecture diagram showing a user request hitting an EC2 Auto Scaling service and getting divided to EC2 instances.

How does auto-scaling work?

Users send requests
The requests are routed to EC2 Auto Scaling service
The service then routes requests to the active EC2 instances
If demand increases, it starts adding new EC2 instances to manage the additional load

An architecture diagram showing multiple user requests hitting an EC2 Auto Scaling service and getting divided to active and backup EC2 instances.

How does auto-scaling work?

Users send requests
The requests are routed to EC2 Auto Scaling service
The service then routes requests to the active EC2 instances
If demand increases, it starts adding new EC2 instances to manage the additional load
As demand goes down, the newly added EC2 instances are shut down

An architecture diagram showing a user request hitting an EC2 Auto Scaling service and getting divided to EC2 instances.

Load balancing vs. auto-scaling

Load balancing

Route traffic evenly
Utilize existing EC2 instances

An icon showing AWS Load Balancer.

Auto-scaling

Ensure demand is always met
Ability to add/remove EC2 instances

An icon showing EC2 Auto Scaling.

Let's practice!

AWS Cloud Technology and Services