Research - Private cloud is cheaper than public cloud…if you can scale it.
- Summary:
- A report released today by 451 Research weighs up the total cost of ownership of private and public clouds, taking into account utilisation and labour efficiencies.
Yes, public cloud can give you the scalability and responsiveness that you require, but if it’s going to cost more, that does make it somewhat harder to justify when putting together a business case.
And cloud pricing is very complex. As well as being dependent upon a huge variety of factors - such as type of deployment, size of deployment, type of workloads, labour costs, utilisation rates - industry pricing is not always entirely transparent.
Which is why the latest report from 451 Research on private vs public cloud pricing may prove interesting for IT buyers.
The results vary, but the general message from the report is that private cloud solutions can work out cheaper than public cloud solutions, if you’re certain of your utilisation rates and labour efficiencies. Equally, Dr. Owen Rogers, Research Director of the Digital Economics Unit at 451 Research, notes that the findings will likely point to why enterprises will operate a number of different models for the time being. He said:
The broader finding is that ultimately end users at enterprises are going to have to use a bunch of models for a bunch of different workload requirements. There might be some workloads that are good at getting that high level of utilisation and are relatively easy to manage, and in those circumstances a private cloud might be better.
But then there might also be some workloads in exactly the same enterprise that are a lot more variable, that are a lot more difficult to manage, and in those cases it may be better to use the public cloud.
The findings
For example, if an organisation is managing its own private cloud and if each engineer at the organisation is managing under 400-500 virtual machines (VMs), with 100% utilisation, public and manage private cloud options are likely to be cheaper. Because of the poor labour efficiency rates.
The report notes, however, if the the primary objective is to be cheaper than the public cloud - then organisations should also take into consideration OpenStack deployments on-premise. It states that it is easier to achieve a low cost using OpenStack than it is using a commercial offering at more than 400-500 VMs per engineer.
For example, to achieve a price point per VM of $0.07 (2 cents cheaper than public cloud), a labour efficiency of just 750 VMs to one engineer is needed for an OpenStack distribution. For a DIY on-premise cloud to be cheaper than public cloud the ratio would need to be 600:1. 451 Research believes that this is uptick in labour efficiency is possible, given the growing maturity of OpenStack tools. It reads:
The question is, will a distribution help an end user increase the number of VMs managed by an engineer from 600 to 750? We think yes, and the benefits of easier installation, patching and management will allow an engineer to manage 33% more in this case.
Private vs public
The report also found that managed private clouds, on average, are going to be more expensive than public clouds. However, at 100% utilisation rate, the premium is “tiny” - just 6%. With a 50% utilisation rate the price jumps to twice the price of public cloud, but 451 Research notes that “some would consider this to be good value considering the benefits of single-tenancy, control, performance and security.
The report states that the best way to achieve private cloud savings is fairly obvious: increase labour efficiencies, utilisation, or both.
At the extremes, at 90% utilization, only a 500:1 labor efficiency is need to break even on private cloud; at 2,000:1 only 40% utilization is needed. A typical midpoint is 60% utilization and 1,000:1 efficiency. Are such levels of utilization and labor possible in practice? Yes, appropriate tools can aid resource allocation, scheduling, automation and simplification. However, for most end users, public cloud is going to be the cheapest option; self-managed private cloud is best for end users that have the funds, desire, labor and tools.
However, as noted above, IT buying decisions aren’t all about predicted cost. 451 Research highlights that there is a “black hole of risk” associated with designing and buying a private cloud solution, particularly when you take into account the buyers may be ambitious about their utilisation and labour efficiency rates. Things also change. So even if you predict those rates accurately at the start of a private cloud lifecycle, they may not remain constant. The report reads:
Let’s consider an extreme example, where an end user has invested millions in a private cloud, and only one VM has been used. That VM is essentially the same price as the whole private cloud. With private cloud there are good gains to be made when utilization and efficiency run high, but there are horrendous losses to be made when they run low.
This risk doesn’t exist with public cloud. End users can turn off resources if they are not being used, and spin them up when they are. On the public cloud, utilization and labor efficiency are the provider’s problem; on the private cloud, the end user takes the risk.
Dr. Owen Rogers reiterated this point to me in our discussion late last week, where he said that private clouds are going to be useful when the risks are minimal, the usage is known and things are not likely to change in the medium term. He said:
If we had a private cloud and it was a very constant capacity - so let’s say it was used for batch processing overnight, and all of the private cloud was used to process that data. Because each workload would essentially be the same, that would mean one engineer would be able to manage a huge amount of them. In such a homogenous environment, all of the virtual machines are being used, so that might be the optimum.
There is very little risk on the enterprise on using public cloud, from an economic perspective, because they don’t have to make commitments, they can grow and shrink as they need. With the private cloud there is an element of risk, because if they fail to achieve a certain level of utilisation and labour efficiency, they’re going to end up paying more.
My take
A useful reminder that whilst cost is a factor in the decision making process, it isn’t the only factor.