The first time you launch an EC2 instance and hit the "Add storage" screen, it's bewildering. gp3? io2? Block Express? It says you can grow to 16TB, but why is io2 expensive and sc1 cheap, and why does instance store throw a warning that "if this instance dies, so does your data"? The answer starts from one fact: EBS is distributed block storage layered over the network. Inside that single word "disk" hide wildly different IOPS, throughput, latency, durability, and replication models — and those trade-offs are the heart of the exam.
Today we follow the trail from EBS's internal architecture (why it's AZ-bound, why only io2 Block Express hits 256K IOPS) down to the physical NVMe of instance store