Teasing apart the public vs. private cloud polemic
Wed 26 Aug 2015
The last few years of debate about the public versus the private cloud have been lively, but arguably not framed by the context of how – and which – industries are driving adoption…
The ambient noise around public vs. private clouds can be confusing to IT managers. After all, the industry has been using terms like private cloud, public cloud, and hybrid cloud for a while and though it seemed like the consensus was that hybrid cloud was going to be the future, the high economic stakes of IT transformation has continued to fuel an ongoing debate.
On the one hand, cloud vendors are doing victory laps and claiming that traditional IT vendors are stonewalling the argument. Articles such as The Private Cloud is for Suckers typify a public cloud-centric narrative.
Meanwhile, the private-cloud camp is pushing back fiercely, claiming that any simplistic rhetoric of a slam-dunk for public cloud is short-sighted and premature. A recent article entitled It’s Time to End the Debate About Private vs. Public Cloud is a good example of this push-back, though it’s unlikely that it will actually end the debate. Likewise, data shows that enterprises continue to buy loads of data center equipment. In fact, data center spending is increasing, led by healthcare and financial sector enterprises as IT organizations reinvest to update and upgrade their equipment.
Wading into all this are the analyst firms. Bellwether enterprise IT analyst firm Gartner recently released a report entitled Internal Private Cloud is Not for Most Mainstream Enterprises. The title itself is thought-provoking.
How can an IT manager make sense of all of these crosscurrents? Here I’ll attempt to tease apart the polemics and give a framework for understanding what’s happening in the private and public cloud market.
Defining terms for the public and private cloud
First of all, let’s get clear on terminology. I will refer to ‘cloud’ as meaning an IT service that is standardized, automated and offers self-service to users via a GUI or API. I will refer to public cloud as meaning cloud services such as Software as a Service (SaaS), Infrastructure as a Service (IaaS) and Platform as a Service (PaaS) offered broadly to the market by a commercial provider on a shared infrastructure and accessed via the Internet. Finally I will refer to private cloud as two sub-categories: externally hosted private cloud and internally-hosted, on-premises private cloud.
Both these forms of private-cloud involve an organization allocating dedicated IT infrastructure. This is an important distinction because criticisms are often launched against private cloud without really defining the category. Most criticism has been aimed at the on-premises version of private cloud, which is important to understand. I will refer to hybrid cloud very simply, as the use of both private and public cloud services in any scenario. That could, for example, mean an IaaS that is on-premises but bursts to public cloud, or just using private and public cloud services for different purposes or applications—a “multi-cloud” approach that matches the reality that most enterprise IT organizations are adopting more than one hypervisor (VMWare, KVM, Hyper-V).
Public cloud has huge momentum
Let’s get one thing perfectly clear – right now public cloud is, quite simply, “killing it”. According to the aforementioned Gartner report, between 2011 and 2014, the total number of public cloud VMs grew from 3% to 20% of all VMs. That’s a factor of 20x. In that same period, growth was flat for VMs in on-premises private clouds. Many, if not most, net new applications are being developed for cloud deployment, and the vast majority of those are being deployed to public cloud. In fact, the same Gartner report forecasts that by 2020 less than 5% of applications will be deployed in on-premises private clouds. If that pans out, private cloud will occupy a relatively small portion of the overall pie. To be fair, that’s still a very large pie.
From an adoption point of view, there are many possible reasons to explain why public cloud seems to be having such VM growth as compared to private cloud. An easy explanation and one that makes sense is that it’s not enterprises that are creating all that public cloud volume VM usage and business; it’s really SMBs—specifically startup software companies.
On the flip side, a lot of those SMB startups are offering SaaS applications consumed by businesses, so in a sense they are contributing to the overall drain of applications away from enterprise data centers. When you consider that there is a SaaS solution in place or brewing for most enterprise applications, this certainly supports Gartner’s projection that applications in five years will overwhelmingly be hosted in a public cloud.
Private clouds bring benefits, but there are still struggles
We’ve examined the notion that much of the growth in public cloud VMs may be coming from outside the enterprise organization. It is however true that there is always going to be a place for private clouds due to privacy, security, regulatory or other compelling business reasons. When you consider that Gartner reckons that 77% of VMs in 2014 were still on-premises and not part of a private cloud, then just extrapolate to all the legacy systems, dedicated x86 servers, physical networking and other specialized IT gear like security appliances and industry-specific IT systems.
There is a lot of stuff on-premises today. It’s naïve to think that this stuff just vaporizes overnight, and it’s self-serving to blithely tell IT departments to “just migrate those applications.” Anyone who has ever been around any business that has had an IT department for at least a decade knows that this is far from a trivial pursuit. Private clouds are not only needed for certain types of applications that will never go into a public cloud. They are also needed to help modernize all that infrastructure and making applications ready to migrate to wherever makes most sense in the coming years.
In terms of numbers, despite the fact that the number of private cloud VMs did not grow as a percentage of total VMs, they did grow by a net factor or 3x. Furthermore, those who implement private cloud notice real benefits. For example, Aberdeen Group reports that 71% of businesses report easier application management and administration since deploying private cloud. And to seemingly confirm these observations, IDC analyst reports show the amount of money spent on cloud infrastructure hardware (servers, storage, and Ethernet switches) will grow at the same rate in private and public cloud for the next five years.
Even Facebook, while not an IaaS, provides a telling anecdote in that they’ve created their own open switch products to drive their massive data centers, rather than buying from Cisco or other commercial networking vendors
On the other hand, multiple Gartner reports observe that there are real challenges for those trying to bring up private clouds. A Gartner report titled “Six Reasons Private Clouds Fail and How to Succeed,” observes that many private cloud initiatives in mainstream enterprises are failing to meet their goals. The reasons are not chiefly technical so much as due to organizational, cultural, and skills issues.
In general, it seems that private cloud initiatives that succeed tend to be undertaken by resource-rich organizations such as those in financial services that can muster huge amounts of capital and attract top engineering talent. Or they are highly motivated by a mission-critical security, compliance or proprietary performance needs, such as for supporting high-speed trading. Other organizations often struggle to realistically assess, scope, resource, and conduct proper client engagement. Gartner reserves its most pointed feedback in the “Internal Private Cloud Is Not for Most Mainstream Enterprises” report by advising that only the most resourced and savvy IT organizations should attempt to build an internal private cloud.
How WAN’s history informs the public/private cloud debate
Where is this all heading? First, there seems to be some contradiction between VM growth and data center spend, so the numbers require some interpretation. Though spending is increasing in both private and public cloud sectors, public cloud vendors tend to buy in much higher volumes than the average enterprise, so they benefit from far better economies of scale. This not only makes basic intuitive sense, but we see that in the case of the largest cloud providers, such as Amazon and Google, they aren’t even purchasing integrated systems. Instead, they’re are acting as their own manufacturers and buying CPUs and other components in bulk from Intel and others. Even Facebook, while not an IaaS, provides a telling anecdote in that they’ve created their own open switch products to drive their massive data centers, rather than buying from Cisco or other commercial networking vendors.
Private clouds will always have a role, but history shows us that it plausibly becomes a minority of infrastructure
Fundamentally, public cloud providers as a whole are purchasing infrastructure in a radically different way from enterprises, and this makes a huge difference in how much infrastructure they get per dollar spent. Aside from capital spending advantages, public cloud vendors are most likely practicing the highest forms of operational efficiency through ruthless, industrial-scale automation. The result is that while hardware spending is equal, public clouds attain higher density of VMs per hardware dollar invested. These capital and operational differences can go a long way in explaining the differential in growth rates between hardware investment spending and sheer VM volume, even in the face of “equivalent” data center hardware purchases.
Now let’s first consider the long view. An interesting parallel could be drawn between public and private infrastructure and the historical wide area network (WAN) space. For a long time, WAN backbones were highly private in how bandwidth and connectivity was managed. Completely segregated, time-sliced leased lines were assumed to be de rigueur for a long time, well into the Internet Protocol (IP) age. Ultimately, when Multi-Protocol Label Switching (MPLS) became convincing enough in terms of privacy and security characteristics, the vast majority of all enterprises, government and even defense organizations shifted to MPLS services running over shared network infrastructures. As a result, a small percentage of IT groups operate their own backbone WAN network today.
There’s no way to absolutely predict the future, but it is certainly plausible that we’ll see the same progression happen with private and public cloud over time. Private clouds will always have a role, but history shows us that it plausibly becomes a minority of infrastructure.
Now let’s consider the short-term to intermediate view. Even if public cloud totally dominates at some point in the future, it’s not going to happen overnight. After all, enterprise WANs took many years to decisively shift to IP and MPLS. Factor in that there’s a ton of on-premises infrastructure still being purchased that will live on for years. Furthermore, internal private clouds have an important evolutionary role in carrying hybrid infrastructure forward into greater agility.
How then should IT invest? Analyst firms from Gartner to IDC and others promote the notion of two-speed or bi-modal IT where investments and operational focus goes on two tracks. One is a largely public cloud-based, DevOps-oriented, disruptive innovation track for new applications, and a more private/on-premises-based track for improving and accelerating legacy applications. Many have argued that this bimodal model is overly simplistic, but whatever model you adopt, the tectonic shifts being driven by cloud technologies require more than a business-as-usual approach.