AWS Application Discovery Service

Migration & Transfer

Collects inventory, utilization, and dependency data from on-premises servers so you can plan a migration.

Verified against AWS documentation · 2026-09-09

Key facts

  1. 01Discovery only — it measures your data center, it never moves a workload
  2. 02Agentless Collector: OVA deployed once into VMware vCenter, any guest OS, ~60-min samples — no running processes, no network dependencies
  3. 03Discovery Agent: installed per server (Windows/Linux, VM or physical), ~15-sec samples, adds running processes and TCP connections — and it's the only source that exports to Athena
  4. 04Everything lands in your Migration Hub home Region — set the home Region first or nothing collects. Quotas: 1,000 active agents, 400 servers per application
  5. 05Reality check: closed to new customers since 7 Nov 2025, superseded by AWS Transform (Migration Hub closed the same day). Exams still test the classic model

Exam triggers

Phrases and scenarios that point here.

  • "We don't know what's running in our data center", "assess before migrating", "map dependencies" → Application Discovery Service
  • Right-size EC2 from real on-prem CPU/RAM usage rather than matching old specs → discovery data feeds the sizing
  • Needs running processes or server-to-server connections → Discovery Agent, not Agentless Collector
  • VMware vCenter, hundreds of VMs, "can't install anything on each host" → Agentless Collector

Not this service

  • Actually moving the servers → AWS Application Migration Service (MGN)
  • One console to track migration progress across tools → AWS Migration Hub
  • A cost/TCO business case for finance → Migration Evaluator

Use cases

  1. 01A retailer with a 400-VM VMware estate and a CMDB nobody trusts — one collector appliance produces a real inventory in days instead of chasing spreadsheets from six teams
  2. 02Right-sizing before lift-and-shift: the boxes were bought for peak 2017 traffic, so actual utilization picks the EC2 family instead of copying old specs 1:1
  3. 03Untangling a legacy app with no owner — agent-collected TCP connections expose the forgotten license server it calls nightly, so it isn't stranded on-prem in wave 1
  4. 04Building migration waves: group servers that talk to each other into one application so they move together, instead of stretching a chatty dependency across the WAN mid-cutover
  5. 05Feeding a cost model — export utilization time-series to Athena or Excel and price the same fleet on EC2 before anyone signs the business case

How it's charged

  1. 01The service itself is free — no charge for the agents, the collector, or the discovery database
  2. 02You pay only for what stores and queries the data: S3 for exports, Athena per TB scanned, Kinesis Data Firehose if streaming
  3. 03Enabling data exploration in Athena auto-creates an S3 bucket and keeps writing agent data — forgetting to switch it off after the project is the classic surprise line item
  4. 04No minimums and no upfront commitment; agents running on EC2 instances add nothing beyond the instance cost
  5. 05The Agentless Collector runs as a VM in your own vCenter, so that compute is your on-prem cost, not an AWS charge

Commonly confused with

AWS Application Migration Service (MGN)Migration & TransferAWS Migration HubMigration & TransferMigration EvaluatorMigration & Transfer

References