--- title: VMware to AWS Migration Paths (2026): EVS vs MGN vs Native Replatform Decision Guide description: For a mid-market software estate (~280 VMs, $340k/yr Broadcom renewal), splitting 60% MGN rehost + 25% EVS relocate + 15% replatform cut parallel-run from 22 to 16 weeks, without forcing one hypervisor strategy. url: https://www.joinhorizon.co/blog/aws-vmware-migration-evs-mgn-native-decision-guide-2026/ datePublished: 2026-07-05T00:00:00.000Z dateModified: 2026-10-30T00:00:00.000Z author: palaniappan-p category: Cloud Architecture tags: aws, vmware-migration, evs, mgn, aws-migration, cloud-migration, architecture --- # VMware to AWS Migration Paths (2026): EVS vs MGN vs Native Replatform Decision Guide > For a mid-market software estate (~280 VMs, $340k/yr Broadcom renewal), splitting 60% MGN rehost + 25% EVS relocate + 15% replatform cut parallel-run from 22 to 16 weeks, without forcing one hypervisor strategy. **Amazon Elastic VMware Service (EVS)** reached **general availability on August 21, 2025** ([AWS announcement](https://aws.amazon.com/about-aws/whats-new/2025/08/aws-general-availability-amazon-elastic-evs/)), with **19 commercial Regions** listed on the [EVS FAQ](https://aws.amazon.com/evs/faqs/) as of mid-2026, including expansions to Singapore, London, Mumbai, Sydney, Paris, and São Paulo through late 2025. **AWS Transform MGN (formerly AWS Application Migration Service)** remains the default **block-level rehost** path to EC2, with a default **150** concurrent replicating source-server quota (increase via AWS Support with a migration plan per [MGN at scale guidance](https://docs.aws.amazon.com/mgn/latest/ug/migration_at_scale.html)). Broadcom's subscription-only VMware licensing (post-2023 acquisition) pushed many enterprises into **$180k to $420k/yr** renewal quotes for mid-market VM counts. That is a business driver. This post is the **VMware-to-AWS path decision guide**: when to **Relocate** on EVS, **Rehost** with MGN, or **Replatform** to native AWS. It is separate from the [EVS product deep-dive](https://www.joinhorizon.co/blog/amazon-elastic-vmware-service-evs/), the [7 Rs overview](https://www.joinhorizon.co/blog/aws-migration-strategy-choose-right-approach/), and a [data-center exit program plan](https://www.joinhorizon.co/blog/data-center-exit-large-scale-aws-migration-program/), though it feeds those documents. Artifacts: [migration path decision matrix](https://www.joinhorizon.co/examples/architecture-blog-2026/vmware-migration-path/migration-path-decision-matrix.md), [wave plan worksheet CSV](https://www.joinhorizon.co/examples/architecture-blog-2026/vmware-migration-path/wave-plan-worksheet.csv). > **Benchmark pattern (not a cited client):** Mid-market **B2B software**, **~280 VMs**, mixed RHEL/Windows, **$340k/yr** Broadcom renewal quote, **22-week** parallel-run target on paper. After tri-path assignment (**60% MGN**, **25% EVS**, **15% replatform**): **16-week** parallel-run achievable. EVS track carried vendor-certified ERP; MGN absorbed stateless tiers; replatform removed **42** VMs from scope entirely. > **Short answer:** Split the estate per VM instead of picking one path. Default to AWS MGN for the long tail, use Amazon EVS only where vSphere is contractual or operational debt, and replatform only where refactor ROI is already funded. In the benchmark pattern below, that tri-path split took a 22-week parallel-run target to 16 weeks. ## Key facts (verified 2026-10-30) - Amazon EVS reached general availability on August 21, 2025. [Source](https://aws.amazon.com/about-aws/whats-new/2025/08/aws-general-availability-amazon-elastic-evs/) - EVS runs VCF 5.2.1 on i4i.metal with license portability (BYOL). [Source](https://aws.amazon.com/about-aws/whats-new/2025/08/aws-general-availability-amazon-elastic-evs/) - The EVS FAQ lists 19 commercial Regions as of mid-2026. [Source](https://aws.amazon.com/evs/faqs/) - MGN has a default 150 concurrent replicating source-server quota, increased via AWS Support. [Source](https://docs.aws.amazon.com/mgn/latest/ug/migration_at_scale.html) ## Three paths, one program office | Path | AWS service | You keep | You change | | -------------- | ---------------- | ---------------------- | ------------------------- | | **Relocate** | Amazon EVS | vSphere ops, HCX, BYOL | Location (AWS bare metal) | | **Rehost** | AWS MGN | App binaries, OS | Hypervisor → EC2 | | **Replatform** | ECS/EKS/RDS/etc. | Business logic | Runtime, often data tier | **Opinionated take:** **Default to MGN for the long tail**, **EVS only where vSphere is contractual or operational debt**, **replatform only where refactor ROI is already funded**. Mandating one path for 280 VMs is how programs miss exit dates _and_ overspend on bare metal. ## Decision tree ``` Vendor cert or ops team only knows vSphere for this VM? YES → EVS (Relocate) — verify Region on EVS FAQ NO → Container-ready or managed-service target within 2 quarters? YES → Replatform backlog (may still MGN-bridge short term) NO → AWS MGN → EC2 (Rehost) ``` Score each VM in [wave-plan-worksheet.csv](https://www.joinhorizon.co/examples/architecture-blog-2026/vmware-migration-path/wave-plan-worksheet.csv). ## Amazon EVS: when relocate wins Per [EVS GA](https://aws.amazon.com/about-aws/whats-new/2025/08/aws-general-availability-amazon-elastic-evs/): - **VCF 5.2.1** on **i4i.metal** with license portability (BYOL) - Full admin access, with the same vCenter model as on-premises - **On-demand, 1-year, and 3-year** consumption options Choose EVS when: - ERP or SCADA vendor supports **vSphere 8 on AWS** but not raw EC2 - Team cannot retrain on EC2 patch cadence before DC lease ends - HCX **vMotion** cutover is required for near-zero downtime on large VMs **When NOT EVS:** Estates under **~30 VMs** (MGN + right-sizing usually cheaper), workloads targeting **EKS in 6 months** (pay bare-metal twice), or Regions **not** on the [EVS FAQ list](https://aws.amazon.com/evs/faqs/). ## AWS MGN: when rehost wins MGN block-replicates to a staging subnet, launches **test** then **cutover** instances. Limits to plan around ([migration at scale](https://docs.aws.amazon.com/mgn/latest/ug/migration_at_scale.html)): - Default **150** max active replicating servers (request increase with justification) - Up to **200** source servers per replication job - **EC2/EBS quotas** for replication servers + snapshots, so plan waves Choose MGN when: - VMs are general-purpose Linux/Windows without vSphere-only tooling - You want **native EC2** integration (IAM instance profiles, Systems Manager, Graviton) - Post-migration path includes **right-sizing** and **Graviton** adoption > **What broke:** Week 3 of a single-path MGN program. **Tier-0 ERP** failed cutover: vendor support rejected EC2 hypervisor for certified stack; **36-hour** rollback via MGN reverse replication. **Detection:** vendor ticket + failed smoke test. **Fix:** moved ERP + 14 dependent VMs to **EVS track**; MGN continued for **180** stateless VMs. **Lesson:** assign path **per VM** before calendar commit; see worksheet column `assigned_path`. ## Native replatform: when to skip the VM entirely Replatform **15 to 20%** of a typical estate first because it removes VMs from EVS/MGN scope: - Containerized microservices already on VMs → **ECS/Fargate** or **EKS** - Commercial databases on VM → **Aurora/RDS** (see [managed DB ops guide](https://www.joinhorizon.co/blog/aws-managed-database-ops-rds-aurora-buyer-guide-2026/)) - Batch → **Lambda + Step Functions** where runtime allows Do **not** replatform tier-0 monoliths on the critical path to DC exit unless refactor is funded and tested; [scale-in-place before decomposition](https://www.joinhorizon.co/blog/aws-legacy-monolith-scale-in-place-before-decomposition-2026/) applies. ## Wave planning: parallel-run math | Wave | Contents | Typical duration | | ---- | ------------------------------------- | ---------------- | | 0 | Discovery + path assignment worksheet | 2 weeks | | 1 | MGN tier-2/3 (low dependency) | 3–4 weeks | | 2 | EVS cluster + HCX tier-0 | 4–6 weeks | | 3 | Replatform pilots (parallel) | 6–12 weeks | | 4 | Retire on-prem networking | 1–2 weeks | Model your VM list in [wave-plan-worksheet.csv](https://www.joinhorizon.co/examples/architecture-blog-2026/vmware-migration-path/wave-plan-worksheet.csv): `tier0` VMs with `vmware_only_tooling=yes` should not default to MGN. ## What to do this week 1. **Export VM inventory** (name, OS, vCPU, RAM, storage, tier) into [wave-plan-worksheet.csv](https://www.joinhorizon.co/examples/architecture-blog-2026/vmware-migration-path/wave-plan-worksheet.csv). 2. **Mark vSphere-only systems.** Vendor-certified vSphere-only systems are candidate **EVS**. 3. **Score the top spend.** Run [decision matrix](https://www.joinhorizon.co/examples/architecture-blog-2026/vmware-migration-path/migration-path-decision-matrix.md) scoring for the top **20%** spend VMs first. 4. **Open a quota ticket.** Open an AWS Support ticket for **MGN quota** if concurrent replicate count exceeds **120**. 5. **Verify EVS Region.** Check **EVS Region** availability for primary + DR before signing relocate SOW. > **Reproduce this:** Download [wave-plan-worksheet.csv](https://www.joinhorizon.co/examples/architecture-blog-2026/vmware-migration-path/wave-plan-worksheet.csv). Fill `assigned_path` for every row. Count paths: if **EVS > 40%** of VMs, validate bare-metal budget; if **MGN > 80%**, audit for hidden vSphere dependencies before cutover calendar. ## What this post doesn't cover - **VMware Cloud on AWS (legacy VMC)** commercial terms: see [EVS vs VMC comparison](https://www.joinhorizon.co/blog/amazon-elastic-vmware-service-evs/). - **MAP funding math:** [MAP SMB guide](https://www.joinhorizon.co/blog/aws-migration-acceleration-program-map-smb-guide/). - **Post-migration FinOps:** [FinOps handoff 2026](https://www.joinhorizon.co/blog/aws-post-migration-optimization-finops-handoff-2026/). - **Application dependency discovery tooling:** covered in [migration readiness checklist](https://www.joinhorizon.co/blog/aws-cloud-migration-readiness-assessment-checklist/). We have not benchmarked EVS **i4i.metal** unit economics against MGN **Graviton** targets for the same workload; fill your SKUs in the worksheet before executive cost sign-off. ## Sources - [Amazon EVS general availability announcement](https://aws.amazon.com/about-aws/whats-new/2025/08/aws-general-availability-amazon-elastic-evs/) (AWS) - [Amazon EVS FAQ](https://aws.amazon.com/evs/faqs/) (AWS) - [AWS MGN migration at scale](https://docs.aws.amazon.com/mgn/latest/ug/migration_at_scale.html) (AWS documentation) **Related:** [AWS migration services](https://www.joinhorizon.co/services/aws-migration/) · [Architecture review](https://www.joinhorizon.co/services/aws-architecture-review/) ## FAQ ### When should we choose Amazon EVS over AWS MGN? Choose EVS when vSphere operations, HCX live migration, or VMware-only vendor certifications block a hypervisor change before your data-center exit deadline. Choose MGN when you can run on EC2 with block-level replication and want native AWS integration without bare-metal VMware licensing overhead. ### When should we NOT use EVS for the whole estate? Skip EVS-only strategies for small estates under ~30 VMs where MGN to right-sized EC2 is cheaper, when your target Region lacks EVS (verify the AWS EVS FAQ list), or when your 6-month goal is EKS/ECS-first. Relocate only the VMs that truly require vSphere. ### What breaks during MGN cutover at scale? EC2 and EBS quotas throttle staging replication; default MGN concurrent replicating server limit (150, increase via Support) stalls waves above ~120 simultaneous sources. Symptom: replication lag grows while cutover calendar slips. Fix: wave planning, quota pre-checks, and staggered cutover instead of a single big-bang weekend. ### How does this differ from the Amazon EVS product post? The EVS post explains vSphere-on-AWS mechanics and BYOL. This post is the portfolio decision: which VMs take EVS, which take MGN, which replatform, and in what order, for mixed estates exiting on-premises VMware. ### Can we use EVS and MGN in the same program? Yes, and it is recommended for most mid-market estates. EVS for VMware-bound tier-0 systems; MGN for general-purpose Linux/Windows that tolerate EC2; replatform for container-ready or SaaS workloads. One program office, three technical tracks. ### What could go wrong with a single-path mandate? EVS-only delays container modernization and bare-metal cost; MGN-only breaks vendor certs tied to vSphere; replatform-only misses the DC exit date. The wave worksheet forces per-VM assignment instead of executive "pick one R." --- *Source: https://www.joinhorizon.co/blog/aws-vmware-migration-evs-mgn-native-decision-guide-2026/*