Guide

RVTools to AWS and Azure cost estimate: what your export can and cannot tell you

Published November 2, 2026 · CloudTiebreak

An RVTools export is the quickest way to describe a VMware environment to anyone pricing a move to AWS or Azure. It is a good starting point for a cost estimate, and only a starting point. This guide shows how to export, which data drives the price, how the same VM can produce very different numbers, and what an inventory cannot tell you.

RVTools tells you what you have. A cloud estimate tells you what that configuration might cost. A business case tells you whether moving it makes financial sense.

What RVTools is

RVTools is a free Windows application that connects to vCenter Server or an ESXi host and lists what it finds: virtual machines, hosts, datastores, networks, snapshots and more. Exporting an inventory only reads from vCenter. RVTools is now distributed by Dell, and its export is accepted as an input by the migration tools of AWS, Microsoft and Google.

How to export your inventory

1. Download RVTools from the official source. Dell’s support page is the authorised download location; the current version is 4.8.2. Microsoft’s documentation notes a supply-chain attack that put malware into the RVTools installer in May 2025, and recommends downloading only from official sites and checking the file hash.
2. Connect with a read-only vCenter account. RVTools does not need administrator rights. A read-only role on the whole vCenter is enough to see every VM.
3. From the File menu, choose “Export all to Excel”. You get one .xlsx file with a tab per inventory type. If you have several vCenters, export each one.

vInfo is enough to start, not to finish

A simplified estimate of compute and storage can be built from the vInfo tab alone: one row per VM, with the columns below.

A migration assessment needs more than that. The other tabs carry the detail that changes the design and, often, the cost: each virtual disk and its controller (vDisk), space used inside the guest operating system (vPartition), host CPU models and a snapshot of utilization (vHost), networks and addresses (vNetwork), snapshots that would be copied or dropped (vSnapshot), and memory and CPU reservations (vMemory, vCPU).

The providers’ own tools reflect this:

• Azure Migrate imports the .xlsx export (as a preview feature, for fewer than 20,000 servers) and reads vHost, vDatastore, vSnapshot, vPartition, vMemory, vDisk and network tabs as well as vInfo. Microsoft Learn
• Google Migration Center requires six tabs for an .xlsx import: vInfo, vCPU, vDisk, vPartition, vNetwork and vHost, from RVTools 4.3.1 or later. A CSV import needs only vInfo. Google Cloud
• AWS Transform accepts the Excel file or a ZIP of its CSV files. AWS guidance asks for at least the vInfo sheet and recommends including vCPU, vMemory, vDisk, vHost, vCluster and vNetwork. AWS docs, AWS blog

So keep the full export. Use vInfo for a first look at cost; send the complete file when someone assesses the migration itself.

One VM, four different answers

Take one VM as RVTools reports it:

Here is what it costs per month on each cloud under four sets of assumptions. Every figure is a public on-demand list price, so you can reproduce it.

On-demand list prices on November 2, 2026, from the AWS Price List API and the Azure Retail Prices API, 730 hours a month, in US dollars. Compute plus one SSD disk only: no backup, snapshots, network, support or taxes. Scenario D excludes the cost of the licences and their Software Assurance or subscription.

The disk: the same 500 GiB, priced two ways

In scenario A, compute costs exactly the same on both clouds. The whole difference comes from storage. AWS gp3 is billed per GB, with 3,000 IOPS and 125 MiB/s included at any size. Azure Premium SSD is billed by size tier, so 500 GiB pays for the 512 GiB P20 tier, and its performance comes with the tier: 2,300 IOPS for P20, 1,100 for P15.

Scenario B assumes the disk can shrink to 200 GiB: the 120 GiB in use plus room to grow. That is a decision to validate, not a setting. “In Use” in vInfo is space on the datastore; the space used inside the guest is in vPartition. Migration tools generally create target disks at the source size, and an EBS volume cannot be made smaller in place, so resizing is planned work. On Azure Premium SSD it also lowers the disk’s performance.

The instance family

Moving from Intel to AMD instances lowers compute by about 5% on both clouds here ($28.03 a month on AWS, $29.20 on Azure). Intel and AMD do not perform identically per vCPU, so the cheaper family is a choice to test with the workload, not a free saving. If two estimates disagree, ask which instance families each one used.

The Windows licence

For this VM, the Windows Server licence is $268.64 of the $548.96 compute price: almost half. What you can do with licences you already own differs by cloud. On Azure, Azure Hybrid Benefit lets customers with active Software Assurance or qualifying subscription licences pay the base compute rate, with a minimum of 8 core licences per VM. On AWS, Windows Server cannot be brought to shared instances; bringing your own licence requires Dedicated Hosts and licences acquired before October 1, 2019. Neither option is free: licences and Software Assurance have their own cost, which belongs in the business case. SQL Server and other licensed software inside the VM do not appear in RVTools at all.

Six things that distort an estimate built from RVTools

1. Powered-off VMs and templates counted as servers

A meaningful share of an inventory is often switched off. If those rows are priced, the cloud looks more expensive than it will be. Filter on Powerstate “poweredOn” and Template “False”, then decide separately which powered-off VMs really need to move.

2. MiB read as MB

RVTools reports memory and disk in MiB. 1 GiB is 1024 MiB, not 1000. A VM with 8192 MiB that becomes 8.19 GB can be pushed to the next instance size if a tool rounds up. Check that 8192 MiB becomes 8 GiB.

3. No statement of which disk figure was used

Total capacity, provisioned and in use can differ by a factor of four on a thin-provisioned VM, as in the example above. A careful estimate says which one it used and why.

4. Allocated size treated as the right size

vInfo shows what each VM has been given, not what it needs. Rightsizing is not simply shrinking: a VM can need a different size because of processor architecture, peak behaviour, memory requirements, network bandwidth limits or disk performance limits of the target instance. Getting it right takes performance history over weeks, from vCenter performance data or from a collector such as the Azure Migrate appliance or AWS Migration Evaluator. RVTools gives a point-in-time view only.

5. Processor family chosen without saying so

As the example shows, the same vCPU and memory cost different amounts on Intel and AMD instances. An estimate should name the families it used.

6. Licences left implicit

Windows licence included or brought, SQL Server or not: these can move the total more than the servers themselves. Mark which VMs run licensed databases before anyone prices the move.

What RVTools cannot tell you about cloud costs

Financial assessments of a migration come in three levels. Each one answers a different question, and each needs more data than the one before.

A well-bounded level 1 estimate is useful: it shows the order of magnitude, the gap between providers and which assumptions matter most. It is not a business case, and nobody should sign a migration contract on it alone.

What is in the file before you share it

An RVTools export describes your environment in detail: VM names, hostnames, IP addresses, annotations, folder paths, cluster and datastore names. Before you send it to a vendor or partner, decide whether they need all of it. You can delete tabs and columns they do not need, or replace names with codes. AWS has published a local tool that anonymizes RVTools exports while keeping the relationships between hosts, clusters and networks.

How the CloudTiebreak estimate works

Our free estimate is a level 1 estimate. These are its assumptions, applied the same way to both providers:

• It reads the vInfo tab, skips powered-off VMs and templates, and converts MiB at 1024 per GiB.
• Each VM is matched to the cheapest instance with at least the same vCPU and memory, without rightsizing, on the processor you choose (Intel, AMD or the cheaper of both). The families compared are x86 families from the two latest Intel and AMD generations on each cloud, where a vCPU is a hardware thread as in VMware; on AWS that leaves out the newer AMD families, whose vCPUs are full cores.
• Disk size comes from Total disk capacity, or Provisioned, or In Use, in that order, and the result says which one was used. Each VM’s disk is priced as one SSD disk: gp3 on AWS, Premium SSD on Azure.
• Prices are public on-demand list prices for the region you pick, refreshed every week from the AWS Price List API and the Azure Retail Prices API. Windows includes the licence; SQL Server and other software do not.
• It leaves out data transfer, backup, monitoring, support, networking and the migration itself, and says so next to the result.

Your file stays in your browser. The page reads the export with JavaScript on your computer; the file is not uploaded. We record an anonymous event with totals only (number of servers and vCPUs, region, processor choice and the monthly results) to see how the tool is used. No VM names, addresses or file contents. You can check it: open your browser’s developer tools, go to the Network tab and drop the file. The only request you will see is that small event.

Questions

How accurate is a cost estimate from RVTools?

As accurate as its assumptions. For the configuration it describes, list prices are exact; the uncertainty is in what the configuration will become in the cloud: disk sizes, instance sizes after rightsizing, licences and everything outside compute and storage. Treat it as a level 1 estimate.

Should I use provisioned or in-use disk space?

Start with total disk capacity, because that is what the cloud disks will be created with unless you resize them. Use the in-use figure only for VMs whose disks you have checked inside the guest and plan to shrink.

Does RVTools include CPU and memory utilization?

Only a point-in-time view in some tabs, such as host CPU and memory usage in vHost. There is no history, so it cannot show peaks. Rightsizing needs performance data collected over weeks.

Can I use the same export for Azure Migrate, AWS Transform or Google Migration Center?

Yes. Each one imports an RVTools export, with its own requirements on tabs and versions, and each one prices only its own cloud.

Checklist before you trust an RVTools-based estimate

1. Only powered-on VMs priced; templates excluded.
2. Memory and disk converted from MiB at 1024 per GiB.
3. Disk figure stated: total capacity, provisioned or in use.
4. Sizing basis stated: allocated, or measured utilization.
5. Instance families and processor (Intel or AMD) stated.
6. Windows and database licences handled explicitly.
7. Estimate level stated: inventory, infrastructure or business case.

Sources

• Dell: RVTools 4.8.2 installer
• Microsoft Learn: Import VMware servers using an RVTools XLSX file
• Google Cloud: Import RVTools data into Migration Center
• AWS Transform: Discover source data
• AWS: Anonymizing RVTools data for AWS migration planning
• AWS: General Purpose SSD (gp3) volumes
• Microsoft Learn: Azure managed disk types
• AWS Prescriptive Guidance: Bring licenses for Windows and SQL Server workloads
• Microsoft Learn: Azure Hybrid Benefit for Windows Server
• AWS: EC2 On-Demand pricing and Azure: Windows Virtual Machines pricing