EC2 - Elastic Compute Cloud

What is AWS EC2 (Elastic Compute Cloud)?

EC2 (Elastic Compute Cloud) is AWS's virtual machine service. Each EC2 instance is a virtual server running on AWS hardware. You select the instance type (defining CPU, RAM, network, storage), the AMI (Amazon Machine Image - your OS + software), and the region.

Instance Types

Instance types follow the naming pattern family.size - e.g., t3.medium, m6i.xlarge, c7g.2xlarge.

Family	Optimised For	Example Use
t (burstable)	Low baseline CPU with burst	Dev/test, low-traffic web
m (general)	Balanced CPU/RAM	Application servers
c (compute)	High CPU	Encoding, ML inference, HPC
r (memory)	High RAM	In-memory DBs, analytics
i (storage)	High IOPS NVMe	Databases, caches
p/g (GPU)	GPU compute	ML training, rendering
x (extra memory)	Extreme RAM (up to 24 TB)	SAP HANA, in-memory analytics

Sizes within a family: nano -> micro -> small -> medium -> large -> xlarge -> 2xlarge -> ... -> 48xlarge.

Purchasing Options

Option	Description	Savings vs On-Demand
On-Demand	Pay per second, no commitment	Baseline (0%)
Reserved (1 or 3 yr)	Commit to instance family + region	Up to 72%
Savings Plans	Flexible reserved - commit to $/hr spend	Up to 66%
Spot	Spare AWS capacity, interruptible (2 min notice)	Up to 90%
Dedicated Host	Physical server for your use	For licensing compliance

Spot interruption: AWS reclaims spot capacity with 2 minutes notice. Use for stateless, fault-tolerant workloads - batch jobs, CI runners, ML training checkpoints.

Storage

• Instance store: Ephemeral NVMe directly on the host. Extremely fast. Lost when instance stops.
• EBS (Elastic Block Store): Persistent network-attached volumes. Survives stop/start. Multiple volume types (gp3, io2, st1, sc1).
• EFS: Shared NFS file system mountable across many instances.

AMIs (Amazon Machine Images)

AMIs snapshot the root volume + launch config. You can use AWS-provided AMIs (Amazon Linux 2/2023, Ubuntu, Windows), third-party marketplace AMIs, or custom AMIs you create from existing instances (for pre-baked environments).

Auto Scaling Groups (ASG)

ASG manages a fleet of EC2 instances:

• Define min/desired/max instance count
• Launch template specifies instance type, AMI, user data
• Scaling policies: target tracking (maintain 60% CPU), step scaling, scheduled
• Health checks: EC2 status checks or ELB health checks replace unhealthy instances

ASG is the primary EC2 horizontal scaling mechanism.

Placement Groups

Type	Behaviour	Use For
Cluster	All instances in same rack	Low-latency HPC, 10Gbps between instances
Spread	Instances on separate hardware	Critical instances needing fault isolation
Partition	Groups in separate racks	Distributed systems (Kafka, Cassandra)

Key Networking

• Each instance gets a private IP within its VPC subnet
• Public IP (if configured) is ephemeral - changes on stop/start. Use Elastic IP for a static public address.
• Security groups = stateful firewall at the instance level (allow rules only)
• NACLs = stateless firewall at the subnet level (allow + deny rules)

Interview Angle

EC2 questions often focus on cost optimisation (on-demand vs reserved vs spot), instance family selection (right-size for the workload), and scaling patterns (ASG + ALB). For stateless services, spot + ASG gives up to 90% savings. For databases, reserved or savings plans. Always mention that t3 instances have CPU credits - sustained high CPU workloads should use m or c family instead.