Amid all of the buzz around cloud technology, it is critical for IT organizations to understand the economic aspects of this technology beyond differences between CapEx and OpEx that have been discussed ad nauseam.
In this recent post, I made the point that comparing cloud to dedicated or physical servers is akin to comparing buying versus renting a car. The fundamental idea in this analogy is that there is a role for every technology, and while you can force fit a use case to a technology or business model, you will get inferior results.
The public cloud is optimal for new projects, for projects where bursting is needed either because the workload is variable in nature, because it is unpredictable or because it needs to scale fast. Yet, at scale, for stable workloads, dedicated servers clearly have the edge. And then, at very large scale, a private cloud sometimes is more economically advantageous than either public cloud or dedicated.
This is not merely theoretical. Over the past few years we have seen this pattern emerge quite clearly from our customers. We have seen customers move through this process in multiple directions based on the economics of their specific workloads.
Here is the bottom line: Once an organization is spending somewhere between $15,000 and $25,000 on public cloud alone, there is a good chance they will see performance and economic benefits by moving to dedicated infrastructure – provided the workloads are stable and don’t require cloud-specific services. With a hybrid cloud, these organizations can get the best of both worlds by hosting the stable parts of their workloads on dedicated servers and the bursting or unpredictable components in a public cloud.
Similarly, organizations that spend somewhere around $100,000 a month on dedicated infrastructure should consider a private cloud, as it could give them a significant economic advantage while retaining many of the benefits of public cloud. And again, with a hybrid cloud, these organizations can combine a private cloud, dedicated and public cloud to get the best of all worlds.
The graphic below is a nice representation of a chart I drew on the back of a napkin when having this conversation a few days ago:
There are no absolutes – I am not trying to say the public cloud is not economically viable at scale or that everything will end on a private cloud in the future. There are many other considerations such as performance requirements, security, compliance and support that influence the right choice in infrastructure. Further, with hybrid cloud it is likely that a workload will combine more than one model; for example, dedicated infrastructure for stable aspects of the workload that need to meet compliance requirements or need high performance, and public cloud for more variable aspects of a workload.
Each deployment model has a sweet spot in relation to the scale of an application and it is wise to understand the economic aspects of these technologies when making infrastructure decisions and planning for the future.
At Rackspace we have helped many organizations go through this evolution and we offer advice and best practices to help our customers think about these decisions. Larger IT organizations, for example, can benefit from an application profiling workshop to get an assessment of the right fit for their application.
Many of our customers have gone through this typical evolution: when a company (or a project) is created, it is easy to start small on the cloud. Public cloud has been incredibly transformational for startups (both in Silicon Valley and for teams within large companies). Having instant access to compute and storage, on demand, without having to invest capital is fantastic.
These same organizations may experience challenges once they grow. They may find that scaling their cloud deployments introduces a level of complexity, over-provisioning and spending that is not optimal. HubSpot, to name a specific example, looked at its application portfolio and moved certain aspects of its workloads to private clouds, and some others to dedicated servers while leaving some in the cloud. The end result, according to HubSpot CIO Jim O’Neil, is four times the efficiency for half the cost.
HubSpot is not an exception. It is a case that exemplifies a pattern we see regularly, where IT departments realize that public cloud can be wonderful, but is not necessarily the right technology to support every workload. More and more companies are taking a second look at the alternatives before making technology decisions, and that’s a really good thing.