SimAWSby ShahriarLabs
Search

Virtual machines · Chapter 2 of 2

How EBS provides persistent block storage

Block storage volumes are physically pinned to single data centres to ensure low latency, but understanding snapshots and IOPS limits lets you move and scale compute storage safely.

14 min2 sectionsBack to Virtual machines

FoundationsBuilds the idea from nothing. No prior AWS assumed.

Volume placement inside an Availability Zone

How do you move an EBS volume to a different Availability Zone?

eu-west-1aeu-west-1bEC2 Instancei-0aa11bb22EBS Volumevol-0998877EBS volumes are SAN storage restricted to a single Availability Zone and cannot attach directly to instances in another zone.

An EC2 instance runs in Availability Zone eu-west-1a.

1/5
  1. 01

    The local nature of block storage

    An Elastic Block Store volume is designed like a virtual hard disk. To ensure low latency and high performance, the volume is created inside a specific Availability Zone. It can only be attached to EC2 instances that live in that same zone.

  2. 02

    Moving data via snapshots

    To move an EBS volume across Availability Zone boundaries, you must take a snapshot. Snapshots are stored in Amazon S3, making them region-wide resources. You can then restore a new EBS volume from that snapshot inside your target Availability Zone.

  3. 03

    Incremental snapshots keep costs down

    EBS snapshots are incremental. The first snapshot copies the entire volume; subsequent snapshots only copy the blocks that changed since the last backup. This design reduces both the time required to back up and the S3 storage fee.

Check yourself

You have an EC2 instance in us-east-1a and an EBS volume in us-east-1b. How can you attach the volume to the instance?

Under the hoodThe same thing from underneath: limits, failure modes, numbers.

Provisioning IOPS and volume performance

How do gp3 and io2 volume specifications impact application performance?

eu-west-1aeu-west-1bEC2 Instancei-0aa11bb22EBS Volumevol-0998877EBS volumes are SAN storage restricted to a single Availability Zone and cannot attach directly to instances in another zone.

An EC2 instance runs in Availability Zone eu-west-1a.

1/5
  1. 01

    The baseline performance of gp3

    General Purpose SSD volumes provide a baseline performance that is independent of the volume size. You get a set amount of IOPS and throughput included in the base price, and you can pay to provision additional performance without buying more storage.

  2. 02

    Provisioned IOPS for intense workloads

    For database engines that require consistent, high-performance block storage, Provisioned IOPS volumes are used. These volumes allow you to configure very high limits, ensuring the storage layer does not become a bottleneck during peak traffic.

  3. 03

    EBS-optimised instances prevent bottlenecks

    EBS traffic travels over a dedicated network link rather than the standard instance network interface. Using an EBS-optimised instance type ensures this storage traffic does not compete with your application web traffic, preventing network contention.

The numbers

gp3 baseline performance
3,000 IOPS and 125 MiB/sIncluded in the storage price regardless of the volume size.
gp3 maximum performance
80,000 IOPS and 2,000 MiB/sRegional gp3 limits depend on volume size and provisioned IOPS. Outposts has lower limits.
io2 Block Express maximum IOPS
256,000 IOPSDesigned for the most demanding database workloads.
Maximum volume size
64 TiB for regional gp3 and io2; 16 TiB for gp2A volume cannot be shrunk once grown, so oversizing is a permanent cost rather than a reversible mistake.

Check yourself

What performance advantage does gp3 offer over the older gp2 volume type?

Official references & further reading

These lessons simplify selected behaviors for learning. Verify current service limits, Region support and production requirements with the official references. Experiments describe their own assumptions.

Report an error or suggest a clearer explanation →