It’s 2024, we know that the vast majority of organizations have adopted either a public cloud or private cloud. Within these organizations, we’ve noticed a significant challenge for IT professionals is to determine the right placement of their dollars against cloud services provided by either public clouds or private clouds.
To help with this, we took the time to layout a cost map and cost/scale guide to assist these professionals. It is important to us that IT professionals understand they do have options and what key driving forces should be taken into account when making a decision.
This analysis can be quite complicated, please bear with the evolution of this public cloud vs private cloud cost tipping point – it is changing over time.
To convince you of the importance, I’m jumping right into the numbers, then we’ll get back to the process.
2024 Total Monthly Cost of VM with Block Storage and Bandwidth – Public Cloud vs Managed Private Cloud Pricing
| Deployment Size, Bandwidth | Public Cloud | Private Cloud | Managed Private* | Public vs Managed | Yearly Diff |
| Small – 100VMs, 10TB | $7,731 | $1,952 | $2,855 | $4,876 | $58,512 |
| Medium – 500VMs, 50TB | $28,925 | $7,375 | $11,294 | $17,631 | $211,575 |
| Large – 1000VMs, 150TB | $60,450 | $11,533 | $16,288 | $44,163 | $529,950 |
| XL – 2000VMs, 300TB | $86,260 | $18,570 | $24,775 | $61,485 | $737,820 |
| XXL – 3000VMs, 600TB | $123,061 | $25,770 | $33,788 | $89,274 | $1,071,282 |
| XXL – 4000VMs, 1200TB | $172,571 | $33,742 | $43,898 | $128,673 | $1,544,077 |
Did the numbers catch your eye? We hope so! You can calculate private cloud pricing yourself using our Cloud Deployment Calculator.
Note this has been updated from the original article as deployment sizes have grown and public and private cloud costs have changed. Here is a screencap of the earlier data. Note the cost of traditional public cloud has grown but also so are the size of workloads we are seeing come over.
Please Note: Each brand is a registered trademark of the company or entity listed. Information presented here was collected in March 2021 and May 2024. If you see any discrepancies or material you feel is inaccurate, please let us know.
Company or Product Names = Key
- Amazon Web Services = AWS
- Microsoft Public Cloud = Azure
- Google Cloud Platform = GCP
- OpenStack with Core Projects = OpenStack
- OpenMetal Private Cloud Core = OpenMetal
- VMWare vSphere plus Hardware = VMWare
- Nutanix Base plus Hardware Vendor = Nutanix
Terms
Public Cloud = Large infrastructure providers that sell you usage based resources one small unit at a time.
Virtual Private Cloud = A software construct that is used to create private networks so resources, like VMs, can communicate easily with each other.
Hosted Private Cloud = A small to large footprint cloud leased to you by a 3rd party in their data center. Learn more if new to you.
Private Cloud = A small to very large footprint cloud where you may own or lease the hardware directly and place it in your data center or chosen colocation provider. Often referred to as an on-premise private cloud but colocation is more common.
Cost and Scale Transitions for AWS, Azure, GCP, OpenMetal, OpenStack, VMWare, and Nutanix
To help compare across public and private clouds by service and by cost, there are a few critical items that will swing your ROI evaluation of the options. First and foremost:
Does your company already have private infrastructure and staff in place?
If you already have staff and data center space, you almost certainly have sunk costs – things that you already pay for that you are not fully using. Space, power, core network capacity, IP space, and connectivity all fit within this. System administrators or DevOps staff are also common team members that will handle cloud administration.
If YES, adding cloud to your existing data center is often much lower cost than previously imagined and thus your tipping point to cloud on-premise – aka private cloud – is lower.
If NO, then your tipping point is going to be higher to move to your own data center-based cloud. When doing that, you are also likely to use managed services from the private cloud vendor until you reach a scale that lets you bring management in-house.
For a NO, most likely you will first go to “hosted private cloud” before going to your own data center-based cloud. The tipping point to go to hosted private cloud is between public cloud and on-premise cloud. If you are well past the tipping point to hosted private cloud, then you will probably go to colocation in a DC. But you will save huge amounts, so there is that.
Know your public cloud to private cloud tipping point
Absolutely critical in your evaluation is understanding the tipping points for certain types of resources.
The tipping points used in the charts below are organized around when public cloud becomes more expensive than alternative clouds. Public cloud, due to its minimum starting point of “1 small unit” is brilliant to get started, but becomes significantly expensive at a certain scale. The famous VC company Andressen-Horowitz published the The Cost of Cloud, a Trillion Dollar Paradox that talks about the loss of huge value for a set of companies caused by the high cost of public cloud.
That concept, boiled far down, has to do with when the cost you pay for the ease of use of the public cloud is overridden by the margin those companies are taking per VM, storage, or other resource. Their margins are very large and so your expense becomes very large once past a tipping point.
For example, using a single AWS, Azure, or GCP VM is quite economical since the expense of all the other things that go on around hardware is simply baked into the cost. But as your deployment grows, the high cost of the individual VM on AWS, Azure, or GCP versus a private cloud VM leaves enough dollars on the table to pay for all the parts around that VM.
To illustrate these, let’s compare using an AWS VM compared to an unmanaged private cloud VM, building in discounts from AWS that you may or may not get. But here is how this happens:
Public Cloud VM Costs vs Private Cloud VM Costs Tipping Point
Public Cloud VM
This was based on the AWS compute-optimized VM with standard performance storage – c7i.large – Xeon-based 2vCPU/4GB plus 30GB standard SSD Elastic Block Storage with 1 weekly snapshot with 1GB diff:
- No commitment – $69.31/month
- 1 year commit – $51.05/month
- 3 year commit – $35.43/month
For small deployments we used the uncommitted, then 1 year, then 3 year. For very large, we also then reduced the per VM as there are confirmed discounts we know customers have received (but not guaranteed) at scale. We used 10% and 5% drop. Public cloud is managed under the VM.
Previously this was based on the AWS VM standard, standard performance storage – t3.medium – Xeon-based 2vCPU/4GB plus 30GB standard SSD Elastic Block Storage with 2 snapshot = $39.36/month, no commitment. With year commit, $30.89/month.
Private Cloud VM
This was based on the OpenMetal Hosted, V3 hardware, VM premium, NMVe network storage – r4.medium – Xeon-based 2vCPU/4GB plus 30GB NVMe SSD Network Block Storage with snapshots. Previously this was based on the V2 version.
For each deployments we based it on the Private Cloud Core (PCC) that was closest in capacity with agreement lengths:
- No commitment, Medium PCC – $15.77/month or $19.24 managed
- 1 year commit, Large PCC – $14.00/month or $17.47 managed
- 3 year commit, XL PCC – $10.97/month or $12.58 managed
- 4 year commit, XL PCC + 5 XL working nodes at 85% – $8.91/month or $9.61 managed
- 5 year commit, XL PCC + 10 XL working nodes at 85% – $8.09/month or $8.61 managed
We provide both private cloud pricing when self managed under the VM and then costs with managed. Management must be either handled by you or contracted from the cloud provider. Private cloud costs for management can be significant per VM when the clusters are small, but reduce to very small amounts per VM when at scale.
VM Monthly Cost Tipping Point by Number of VMs, Public Cloud vs Self Managed Private Cloud
| 2vCPU/4GB | Public | Private | Managed Private | Public vs Managed Private | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Units | VM | Disc | Total | VM | Disc | Total | VM | Disc | Total | Cost/month | Yearly | Tipping Point |
| 100 | $69.31 | – | $6,931 | $15.77 | – | $1,577 | $24.80 | – | $2,480 | $4,451 | $53,412 | |
| 500 | $51.05 | 26.3% | $25,525 | $14.00 | 11.2% | $7,000 | $21.84 | 11.9% | $10,919 | $14,606 | $175,275 | Maybe Here |
| 1000 | $51.05 | 0.0% | $51,050 | $10.97 | 21.6% | $10,970 | $15.73 | 28.0% | $15,725 | $35,325 | $423,900 | Likely by Here |
| 2000 | $35.43 | 30.6% | $70,860 | $8.91 | 18.8% | $17,820 | $12.01 | 23.6% | $24,025 | $46,835 | $562,020 | Definitely by Here |
| 3000 | $31.89 | 10.0% | $95,661 | $8.09 | 9.2% | $24,270 | $10.76 | 10.4% | $32,288 | $63,374 | $760,482 | Ouch |
| 4000 | $30.29 | 5.0% | $121,171 | $7.69 | 5.0% | $30,742 | $10.22 | 5.0% | $40,898 | $80,273 | $963,277 | Absurd |
In the above, a few thousand dollars will not cover self management costs. Running a cloud, even a small one, takes an admin 10-20 hours a month plus some support from off-hours staff.
But – and a critical but – a well-architected large cloud does not require a lot more time. Think 40-80 hours a month for 1000 VMs, then scale up a small amount of time per node that is added. It is important to note that 24/7 coverage requires generally 3+ admins but well-architected clouds don’t require emergency alerts when common hardware fails. Redundancy is built in to not only protect the data and workloads, but to protect the admin’s time as well.
Operational Costs – Do I need a new team to run the underlying Cloud?
You may be thinking that you would add a whole team to handle “running the cloud” and thus your private cloud costs would have to pick up that team. Realistically, you are already running hundreds or even thousands of resources. You have a team, there is no question of this already. The question is can your existing team handle running the cloud in addition to running the VM, storage, and other workloads using the cloud.
For OpenMetal, we have seen companies assign up to 2 dedicated administrators to get started, but typically only 1 or someone part time. In all cases so far though, once procedures and training are finished for the operations team, we have not seen a standalone cloud team being used.
Simply put, well architected clouds do not need a dedicated team to run it. If you are going to be pushing the cloud, which may be part of your business model, there are 3 areas to keep in mind that can increase operational cloud costs but give you greater system efficiencies.
- Pushing used disk space above 85% of capacity. As this is your own cloud, there is a finite amount of storage space unless you are ready to add hardware indefinitely. Since that is a bad idea, you will want an administrator watching this and taking action with the rest of your team to reduce storage creep.
- Overprovisioning of RAM. As this is your own cloud, there can be a tenancy to try to push the hardware to its limits. If you decide to you need to do this, which is not unusual, you need to watch for OOM Kills. This indicates you have pushed the hardware past its limits. The more you do this, the more you will need a system administrator to be involved.
- Customization away from standard deployments. The farther a cloud gets from the typical deployment, the more your team will need to be involved. This is both while running and during upgrades.
Public Cloud Bandwidth Costs vs Private Cloud Bandwidth Costs Tipping Point
Another very large expense in the public cloud is bandwidth. Outbound traffic is called egress. It is calculated by GB, usually shown as TB though, of data transferred out of your service. If you have a VM, you are paying for the VM only, not the transfer of data. Egress is paid for separately.
Public Cloud Bandwidth
Azure egress – This is charged at different rates as follows. This is from North America or Europe out to the internet on Azure’s lower cost version, not the “premium global network”. It is per GB per month.
- First 100GB – Free (ignoring this $8 “benefit”)
- Next 10TB – $0.08/GB
- Next 40TB – $0.065/GB
- Next 100TB – $0.06/GB
- Next 350TB – $0.04/GB
Private Cloud Bandwidth
Bandwidth in OpenMetal, like data center connections, are billed at the 95th percentile. Each Medium and larger cluster comes with a 1Gbps egress Port. Assuming 70% of traffic occurs in a peak time we can estimate per GB costs. This means a 1Gbps port can transmit about 155TB per month with that traffic pattern. Additional 1Gbps egress “ports” are $375/month. In the event of bursting above the committed port, the following comes into play:
- <~1TB – $2.02/Mbit or ~$0.05/GB
- ~1TB to ~100TB – $1.84/Mbit or ~$0.046/GB
- ~100TB to ~400TB – $1.38/Mbit or ~$0.034/GB
- >~400TB – $1.15/Mbit or ~$0.029/GB
Monthly Bandwidth Cost Tipping Point by TB of Transfer – Public Cloud vs Private Cloud
| TB | Public Cloud | Private Cloud | Month Diff | Yearly Diff | Tipping Point |
|---|---|---|---|---|---|
| 10 | $800 | $375 | $425 | $5,100 | |
| 50 | $3,400 | $375 | $3,025 | $36,300 | |
| 150 | $9,400 | $563 | $8,838 | $106,050 | Maybe Here |
| 300 | $15,400 | $750 | $14,650 | $175,800 | Likely Here |
| 600 | $27,400 | $1,500 | $25,900 | $310,800 | Ouch |
| 1200 | $51,400 | $3,000 | $48,400 | $580,800 | Absurd |
Combined Monthly Bandwidth and VM Costs for Public Cloud vs Private Cloud vs Managed Private Cloud
To simplify the explanation, we are assuming, for private cloud analysis, using 3 scenarios out of the 6 Cloud Deployment Options:
Scenario 1
You started on public cloud and do not have physical infrastructure. You have costs already associated with system admins running your workloads in your VMs.
Scenario 2/3
You have a fairly typical private infrastructure footprint within a data center. Fairly typical means you have two or more high volume internet connections (think Level3/Lumen, Zayo, GTT, Cogent, etc.) with proper redundant networking hardware, several free cabinets or the equivalent, and 5+kW of unallocated power. 2 – Your own hardware in a colocation facility. 3 – Your own data center.
With that being said, you may find yourself between these, so reviewing both situations may help you.
Scenario 1 – No physical infrastructure
In this scenario we will look at the public cloud versus managed hosted private cloud options. As mentioned above, the reason for this comparison is because the tipping point between public cloud and hosted private cloud is significantly less then going from public cloud to on-premise private cloud. If your company or team was “born in the public cloud” and do not have experience with data centers and hardware then starting with a private cloud and its more simple costs will save you a lot of headache and wasted time.
OpenMetal is highly automated OpenStack and can act as a stepping stone to self-managed on-premise OpenStack if you believe that is the final destination for your cloud.
On to the numbers! And yes, the differences are real – feel free to check the OpenMetal Case Studies section for customers that have taken advantage of these private cloud cost savings already. All are monthly based cloud prices with a yearly difference total.
| Deployment Size, Bandwidth | Public Cloud | Private Cloud | Managed Private* | Public vs Managed | Yearly Diff |
| Small – 100VMs, 10TB | $7,731 | $1,952 | $2,855 | $4,876 | $58,512 |
| Medium – 500VMs, 50TB | $28,925 | $7,375 | $11,294 | $17,631 | $211,575 |
| Large – 1000VMs, 150TB | $60,450 | $11,533 | $16,288 | $44,163 | $529,950 |
| XL – 2000VMs, 300TB | $86,260 | $18,570 | $24,775 | $61,485 | $737,820 |
| XXL – 3000VMs, 600TB | $123,061 | $25,770 | $33,788 | $89,274 | $1,071,282 |
| XXL – 4000VMs, 1200TB | $172,571 | $33,742 | $43,898 | $128,673 | $1,544,077 |
If you are in scenario 1 – which is very common now – this is a video for technical executives that are exploring how to use on-demand private cloud (aka private cloud as a service). It covers quite a few public cloud versus private cloud topics and should give some additional insight. It also covers workloads that you may go after first when moving.
Scenario 2 – Established physical infrastructure in colocation
In this scenario we will look at colocation versus a hosted private cloud. OpenMetal is popular as an alternative to colocation but there are several tipping points to consider when you have colocation already. Here are the basics:
- If you have considerable investment into the hardware, network connections, and processes adding a cloud under the edge routing is typically a good cost choice.
- If your agreement with the colocation provider does not expire for a long time you should make a plan to grow or sunset out.
- You are close or not close to the capacity. If you are running close to the capacity of your space and do not want to invest more, you may want to consider expanding into a hosted private cloud. If you are not running close to the capacity of your space, your best cost choice could be to add a private cloud in your colocation space.
Scenario 3 – Established physical infrastructure, on-premise data center
In this scenario we will look at the public cloud versus the “on-premise” private cloud options. With significant investment in your data center already completed, you have what is called a “sunk cost”. Almost certainly you have the capacity in your data center now and adding additional hardware is typically quite appealing. If for some miraculous reason you have a 98% full data center with a fully used set of internet connections, well, kudos to you! With your savvy approach, you will already know private cloud is a better deal.
If you are not already running a cloud, consider the following:
- Use OpenStack, but get a well-architected cloud from a provider. OpenMetal does not provide this service, but there are many reputable on-premise OpenStack providers.
- If you are going to build your own private cloud, we recommend you learn on ours first. Get a Small or Medium on a month-to-month. See how it is set up, break and fix it, etc.
Have questions? Chat With Us
OpenStack Service Detail and Notes
OpenMetal can provide private cloud pricing that no other provider can currently offer. This is because we leverage the number 1 private cloud system in the world, OpenStack. OpenStack is the 3rd most developed open source project behind Linux and Chromium. It provides cloud services and hardware automation. It is typically referred to as a “private cloud” when its Core Projects are the only Projects installed in an OpenStack.
The Core Projects provide the following:
- OpenStack Nova – Virtual Machine-based compute
- OpenStack Neutron – Private networking
- OpenStack Cinder – Block storage
- OpenStack Horizon – Dashboard for users and some administration
- OpenStack Keystone – User and API access permission management
- OpenStack Glance – VM images are stored here
Test Drive
For eligible organizations, individuals, and Open Source Partners, Private Cloud Cores are free to trial. Apply today to qualify.
Subscribe
Join our community! Subscribe to our newsletter to get the latest company news, product releases, updates from partners, and more.