Hands-on AWS knowledge base for developers Managed Ops
Talk to Us

The Difference Between AWS Outposts and Local Zones (CLI)

Export PDF Print-friendly layout with a copyright watermark; choose Save as PDF in the print dialog
The Difference Between AWS Outposts and Local Zones (CLI)

TL;DR

  • A Local Zone is AWS-owned hardware in a metro near your users, provisioned on demand from the parent Region. An Outpost is an AWS-owned rack standing inside your own building, ordered weeks ahead and paid for on a committed term.
  • Pick Local Zones when the problem is user latency and bursty demand. Pick Outposts when the problem is data residency, a factory floor, or a site that cannot survive an internet round trip to the Region.
  • Most sizing mistakes I see are not about performance — they're about ordering one when the workload needed the other. You can settle it in about 90 seconds with describe-availability-zones and list-outposts.

One script that tells you which one you actually have

Before comparing the two, enumerate them. Both surface through EC2 APIs, but only one of them shows up as a zone group you can opt into. This script is the first thing I run in any new account:

#!/usr/bin/env bash
# inventory.sh - every non-region zone this account can reach
set -euo pipefail
REGION="${REGION:-us-west-2}"

aws ec2 describe-availability-zones 
  --region "$REGION" 
  --all-availability-zones 
  --query 'AvailabilityZones[?ZoneType!=`region`].{Zone:ZoneName,Group:GroupName,Type:ZoneType,OptIn:OptInStatus}' 
  --output table


The `--all-availability-zones` flag matters. Without it you only see zones already enabled in your account, and that is exactly how people conclude "our region doesn't have Local Zones" when there are three sitting there unopted.

Opting a zone group in is idempotent and takes effect in seconds:

```bash
aws ec2 modify-availability-zone-group 
  --group-name us-west-2-lax-1 
  --opt-in-status opted-in 
  --region us-west-2


Then ask the zone what it can actually run, because the answer differs per metro:

```bash
aws ec2 describe-instance-type-offerings 
  --location-type availability-zone 
  --filters "Name=location,Values=us-west-2-lax-1a" 
  --region us-west-2 
  --query 'InstanceTypeOfferings[].InstanceType' 
  --output text | tr 't' 'n' | sort | head -40


On a run last month that returned roughly 70 instance types, heavy on m5, c5, r5 and t3, with no m7i anywhere in the list. If your launch template pins `m7i.xlarge`, this is where you find out — not at 02:40 when the ASG refuses to scale.

Outposts are a different call path entirely:

```bash
aws outposts list-outposts --region us-east-1 
  --query 'Outposts[].{Id:OutpostId,Name:Name,AZ:AvailabilityZone,Status:LifeCycleStatus}' 
  --output table

aws outposts get-outpost-instance-types 
  --outpost-id op-0a1b2c3d4e5f67890 
  --region us-east-1 
  --query 'InstanceTypes' --output text | tr 't' 'n' | sort


If the first command returns an empty table, you have no Outposts in that account, which is the correct answer for the overwhelming majority of SMB accounts I touch.

## What the two outputs actually prove

A Local Zone is an extension of the parent Region's VPC. Subnets you create there belong to your VPC, route through your route tables, and the control plane stays in the Region. Build an EC2 instance in `us-west-2-lax-1a` and its API calls, IAM, CloudWatch and autoscaling all live in `us-west-2`.

An Outpost is also an extension of the Region, but the rack is on your slab, connected over a service link, and it exposes a local gateway so on-prem networks can reach the instances without hairpinning through AWS. The control plane is still regional — lose the service link for long enough and the Outpost goes to a disconnected state where you can't launch anything new.

| Dimension | Local Zones | Outposts |
|---|---|---|
| Hardware location | AWS facility in a metro near users | Your building, AWS-owned rack |
| Site requirements | none from you | power, cooling, rack space, uplink |
| Lead time | minutes (opt in, create subnet) | weeks to months, scheduled with AWS |
| Capacity model | on-demand, per instance | fixed pool you ordered |
| Latency to parent Region | inter-AZ-ish, single-digit ms to metro users | must stay within AWS's ~20 ms round-trip service-link limit |
| Instance types | subset, varies by metro | subset, fixed by rack configuration |
| Local services | EC2, EBS, NLB endpoints (varies by zone) | EC2, EBS, S3 on Outposts, RDS, EKS local clusters, ECS, ELB |
| Billing | usage-based, rates above the parent Region | committed term for capacity plus service link |
| Typical owner | web/app team chasing user latency | ops team with a residency or factory constraint |

The row that ends most arguments is capacity. A Local Zone gives you on-demand instances with normal quotas, so a Black Friday spike is a quota conversation. An Outpost gives you exactly the CPU and memory you ordered, and if the rack is full it stays full.

##

Sources

Leave a Reply

Your email address will not be published. Required fields are marked *