VM Sizing Calculator for PACS, RIS & VNA
Estimate vCPU, RAM, VM disks, IOPS and cluster hosts from your exam volume and user count. Add HA and performance buffers, project five years of growth against your cluster, and compare cloud with on-prem cost.
Workload
Components
Switch on what your PACS, RIS or VNA deployment runs. Open a component to change its sizing assumptions.
Sizing assumptions
Sizing assumptions
Sizing assumptions
Sizing assumptions
Sizing assumptions
Sizing assumptions
Sizing assumptions
Resilience
Your cluster
The hosts you have today, used for host counts and saturation.
Cost
Total vCPU
105
With buffers applied
Total RAM
234 GB
With buffers applied
VM disks
2,685 GB
OS and data disks, not the image archive
Total IOPS
15,038
With buffers applied
Hosts needed
1
2 with N+1 redundancy
Cluster saturates
Beyond Year 5
First year above 85% utilization
VM allocation
Sized per component, before HA and performance buffers. vCPU rounds up to multiples of 2, RAM to multiples of 4 GB.
| Component | vCPU | RAM GB | OS disk GB | Data disk GB | IOPS | Network Mbps |
|---|---|---|---|---|---|---|
| Application Server | 14 | 28 | 120 | 50 | 1,000 | 1,000 |
| Database Server | 18 | 52 | 120 | 500 | 4,000 | 1,000 |
| Web Server | 8 | 16 | 120 | 20 | 450 | 500 |
| Broker / Integration | 8 | 16 | 120 | 30 | 675 | 500 |
| DICOM Server | 14 | 28 | 120 | 200 | 3,000 | 2,000 |
| AI Module (off) | 0 | 0 | 120 | 100 | 0 | 2,000 |
| Reporting Server | 8 | 16 | 120 | 50 | 900 | 500 |
| Total | 70 | 156 | 840 | 950 | 10,025 | 7,500 |
Components switched off add no vCPU, RAM or IOPS, but their disk and network figures still count toward the totals. Set them to 0 under Sizing assumptions to remove them.
From base to final sizing
| Resource | Base | With HA | With performance buffer |
|---|---|---|---|
| vCPU | 70 | 88 | 105 |
| RAM (GB) | 156 | 195 | 234 |
| VM disks (GB) | 1,790 | 2,238 | 2,685 |
| IOPS | 10,025 | 12,531 | 15,038 |
Hosts needed: 1 by vCPU and 1 by RAM, using the larger. That works out at 105 vCPU and 234 GB RAM per host.
5-year growth against your cluster
Your cluster has 512 vCPU and 2,048 GB RAM. Year 0 is today.
| Year | Exams | vCPU | RAM GB | VM disks GB | IOPS | vCPU use | RAM use |
|---|---|---|---|---|---|---|---|
| Year 0 | 250,000 | 105 | 234 | 2,685 | 15,038 | 20.5% | 11.4% |
| Year 1 | 275,000 | 116 | 257 | 2,954 | 16,541 | 22.6% | 12.6% |
| Year 2 | 302,500 | 127 | 283 | 3,249 | 18,195 | 24.8% | 13.8% |
| Year 3 | 332,750 | 140 | 311 | 3,574 | 20,015 | 27.3% | 15.2% |
| Year 4 | 366,025 | 154 | 343 | 3,931 | 22,016 | 30.0% | 16.7% |
| Year 5 | 402,628 | 169 | 377 | 4,324 | 24,218 | 33.0% | 18.4% |
First year above 85%: vCPU Beyond Year 5, RAM Beyond Year 5.
5-year cost: cloud and on-prem
| Cloud | Cost |
|---|---|
| vCPU / month | $3,150 |
| RAM / month | $1,170 |
| VM disks / month | $210 |
| Total / month | $4,530 |
| Total / year | $54,357 |
| 5-year cost | $271,786 |
| On-prem | Cost |
|---|---|
| 1 host | $25,000 |
| Maintenance / year | $4,500 |
| 5-year cost | $47,500 |
On-prem cost is priced on the 1 host needed for capacity, before N+1. With N+1 you would buy 2.
Cloud covers vCPU, RAM and VM disks. On-prem covers hosts and maintenance. Licensing, power, staff and the image archive are not included in either.
Performance buffer comparison
The base totals at different buffers, without HA overhead.
| Resource | No buffer | 10% | 20% | Custom (15%) |
|---|---|---|---|---|
| vCPU | 70 | 77 | 84 | 81 |
| RAM (GB) | 156 | 172 | 187 | 179 |
| IOPS | 10,025 | 11,028 | 12,030 | 11,529 |
Sizing the image archive as well? Use the PACS Storage Growth Calculator.
How the numbers are calculated
Sizing runs per component, and each component rounds up before the totals are added. The buffers then apply to the totals.
- Base demand, per component
- base + (annual exams ÷ 100,000) × per-100k figure
- vCPU adjustments
- base vCPU × (1 + concurrent users ÷ 100), × 1.25 for the database above 500,000 exams
- Distributed deployments
- vCPU, RAM and IOPS × sites connected
- Final sizing
- component totals × (1 + failover buffer, when HA is on) × (1 + performance buffer)
- Hosts needed
- the larger of final vCPU ÷ vCPU per host and final RAM ÷ RAM per host, rounded up
- Growth and saturation
- final × (1 + growth)^year ÷ (per-host capacity × hosts available), saturated above 85%
- Cloud 5-year cost
- ((vCPU × rate) + (RAM × rate) + (disk TB × rate)) × 12 × 5
- On-prem 5-year cost
- (hosts × cost per host) + 5 × (maintenance % × hosts × cost per host)
Common questions
Does this work for PACS, RIS and VNA?
Yes, as a starting estimate. The calculator sizes seven common server roles from your exam volume and user count, and you switch off the roles your system does not run. It is not specific to any vendor, so edit the sizing assumptions to match your vendor's published figures.
How is each virtual machine sized?
Each component starts from a base figure plus an amount per 100,000 exams. vCPU then scales with concurrent users, the database server takes an extra 25% above 500,000 exams, and distributed deployments multiply vCPU, RAM and IOPS by the site count. vCPU rounds up to multiples of 2 and RAM to multiples of 4 GB.
Does the storage figure include my images?
No. It covers the OS and data disks on the virtual machines only. Size the image archive separately with the PACS Storage Growth Calculator.
How is the host count worked out?
The final vCPU and RAM totals are divided by the capacity of one host and rounded up, and the larger result is used. With high availability on, N+1 adds one more host. On-prem cost is priced on the count before N+1, so add one host to your budget if you plan to run N+1.
What does cluster saturation mean here?
Saturation is the first year projected vCPU or RAM goes above 85% of your cluster capacity, using your annual growth rate. Year 0 is today, so "Now" means the cluster is already over the line.
Can I use this instead of vendor sizing?
No. Use it to frame the conversation and to check a vendor sheet for gaps. Final specifications need validating against your vendor and platform, and against measured peak load once the system is live.
This calculator gives planning estimates and does not replace vendor sizing documents or a formal architecture review. Check final specifications against your vendor and platform requirements.
Related Tools & Resources
Healthcare IT Server & VM Sizing
Why vendor sheets are a floor, how to measure real peak load, and the virtualization pitfalls that cause incidents.
PACS Storage Growth Calculator
Project five years of image storage by modality, with capacity saturation and cost.
Healthcare IT Diagram Builder
Draw the architecture you just sized, with PACS, RIS and VNA nodes and protocol-labelled links.