Skip to main content

Battle in the Cloud

By November 10, 2010Article

Is a private cloud really a cloud? Do virtualization and SaaS software “count” as cloud initiatives? Is cloud computing the end of IT?
The cloud computing marketplace is expanding rapidly with customers adopting a variety of technologies, models and strategies to drive business value in their companies. Our “Leaders in the Cloud” study found that the multitude of approaches to cloud success has not only led to a debate over solutions but also to controversy over terminologies.
In this article, Sand Hill Group Cloud Research Director Kamesh Pemmaraju, Appirio CEO Chris Barbin, Eucalyptus Systems CEO Marten Mickos, and Salesforce.com Director of Platform Research Peter Coffee present their perspectives on what “counts” in the rapidly emerging “Battle of the Cloud.”
Private Clouds Makes Business Sense
Kamesh Pemmaraju, Cloud Research Director, Sand Hill Group
Most people believe the promise of cloud computing is to be able to leverage the expertise of a vendor who is delivering computing power at such a high volume that the per-unit cost of management and maintenance goes down. But building private clouds means the internal staff has to learn and run all technologies associated with the cloud.
When you have to pay for all those capital, maintenance, and labor costs, do private clouds really make economic sense? Is there a certain volume at which a private cloud becomes the right decision?
The answer for many companies is “Yes.”
Corporate data centers are not going away. Cloud computing is not the end of IT. The idea that someday we’re going to throw a switch and move everything out to an external cloud is neither a practical nor a realistic scenario – It just isn’t going to happen in the foreseeable future.
Enterprises have billions of dollars invested in IT infrastructure. It will take several years to fully amortize and depreciate those assets. Cloud computing is simply one new part of the overall technology strategy – just one new tool in the CIO’s toolbox. With external clouds, enterprises want to avoid lock-in and deploy their applications and data to one or more clouds – including internal or private clouds – that meet their needs in terms of security, scalability, compliance, cost, performance and latency.
Furthermore, these enterprises are not ready to migrate their most sensitive and business-critical assets to an external cloud under third-party management, particularly if their infrastructure, services, and data are shared with other customers. The lack of control over mission-critical assets and concerns about security and privacy are driving many enterprises to build their own private clouds.
Not surprisingly, our “Leaders in the Cloud” research study uncovered compelling evidence that the major adoption and growth in the cloud space will take place in private and hybrid clouds over the next three years.
In a typical large enterprise today, one finds a heterogeneous mix of platforms, including mainframes, databases, applications and services. Thousands of applications – at least 20-30 percent of which are mission-critical – still run on old-world hardware that is tightly coupled to other applications and infrastructure. Therefore, not all applications will be able to move to the cloud. Enterprise customers will pick and choose applications and their IaaS, PaaS and SaaS (*aaS) vendors based on their internal business needs and technical and architectural constraints thus creating a diverse and heterogeneous cloud environment.
Most large enterprises that are leaning toward internal clouds are first building the foundational capability of highly virtualized infrastructures. What began many years ago as a data center consolidation and virtualization process to achieve improved efficiencies has now become sound preparation for a private cloud computing strategy.
In what is going to become an increasingly “hybrid” cloud computing world, IT leaders cannot afford not to focus on growing the internal virtualization footprint (with an average of only 20% utilization and lots of room for improvement) and then moving on to building internal cloud computing capabilities that can seamlessly interoperate (e.g, bursting out for peak loads) with external cloud services and vendors.
An internal cloud can provide a higher-level management, workload optimization, and automated orchestration service. Other capabilities include interactive service portals, end-user self-service concepts, flex- up and down, return the asset, swap an asset for higher compute, and so on. All this will lead to empowering application engineering and strategic business agility.
Therefore, no matter what the scale of the IT infrastructure is, building internal cloud computing capabilities will lead to improved agility and lowered costs.
In fact, the belief that large cloud vendors are the only groups with dramatic economies-of-scale may well be misplaced. The same cloud technologies, automation tools, sustainability tools, containerized data centers, and other options are also available to large enterprises. With these powerful cloud building blocks, there is no reason why enterprises cannot leverage private clouds to dramatically decrease IT maintenance costs while increasing business innovation and agility, particularly for strategic and mission-critical applications that drive business value.
Private Clouds Don’t Deliver Unique Business Value
Chris Barbin, CEO, Appirio
“Private clouds” are a way for virtualization, hardware and traditional services vendors to stay relevant and on the radar screen of customers who are increasingly asking why they need to be in the business of managing their own data center.
The reality is that private clouds are a rebranding of datacenter virtualization.
Virtualization is a critical enabler of efficiencies and improved service levels in most organizations and has a clear ROI.
However, it has almost nothing to do with cloud computing.
The attributes that virtualization shares with cloud computing are the ability to self-provision and the ability to scale more quickly than with physical hardware.
Thus, at a superficial level, virtualization looks a bit like infrastructure as a service (IaaS), especially services like EC2 and S3.
However, there are three big differences.
First, IaaS providers like Amazon operate at massive scale and higher efficiency than most enterprises, so their unit cost is significantly lower, and they can scale without additional investment.
Second, scale enables IaaS providers to not only attract the best in the industry at solving infrastructure problems, but also invest more in optimizing and enhancing their infrastructure.
Third, IaaS providers can learn from the actions of their entire customer base and improve their platforms over time.
The last point is significant.
If you build your own virtualized datacenter, it does not automatically improve over time.
So, virtualized datacenters have a few superficial similarities to IaaS but are actually quite different.
Moreover, cloud computing is a lot more than IaaS.
In fact, we believe that attaining the true benefits of cloud computing – business agility, innovation that drives the business and user effectiveness – requires the adoption of higher-level cloud services such as platform-as-a-service (PaaS) and software-as-a-service (SaaS). (Read more in our post ”Cloud Computing Savings – Real or Imaginary?”)
And not even the most ardent private cloud advocate can claim that there’s a connection between virtualized datacenters and PaaS or SaaS.
Let’s compare an Infrastructure-as-a-service (IaaS) offering and an internal virtualized environment.
When you buy IaaS, you are paying the vendor their unit cost and a margin.
In most cases the vendor is operating at much greater scale than the enterprise so the unit cost differential should be sufficient to support a vendor margin and still deliver savings.
Of course, there may be cases in which a virtualized datacenter or “private cloud” is economically comparable to an IaaS service.
However, that’s missing the point of what’s possible with cloud computing.
Cloud computing (which includes SaaS, PaaS, AND IaaS), enables IT to shift its role within the enterprise.
Currently, managing infrastructure consumes more than 70 percent of IT efforts, leaving little time for innovation.
Virtualization does not change this.
By freeing IT from managing physical and platform infrastructure, public cloud applications and platforms shift the mix toward innovation and finally enable IT to deliver business impact.
That is the true economic trade-off that every company needs to evaluate.
If your datacenter is close to the size of Amazon’s or Google’s, a virtualized datacenter may provide similar benefits and economics to a commercial IaaS offering.
However, the real question is whether it’s a core competency of your business to run massive datacenters, and whether it’d be better to apply the resources of your company to the things that help you differentiate yourself as a business.
Raw compute capacity is most often not a core differentiator for a business and is rarely where a company’s executive team is focused.
If your team is running massive data centers – virtualized or not – you’re still in the business of manufacturing IT.
If IT wants to become a true business enabler, the right decision is for IT to shift its focus from manufacturing IT to harnessing IT for business innovation.
Cloud Solutions Will Transcend Boundaries
Marten Mickos, CEO, Eucalyptus Systems
Cloud computing is as much about how computing is delivered to its users as it is about the economies of scale.
The promise of cloud computing includes a notion of opaque omnipresence – of computing available at anytime, anywhere and with practically limitless elasticity (i.e. scaling quickly both up and down) without the user knowing or caring about where the computing power is located.
We use this model today to deliver the common carrier Internet to users: users expect connectivity and have little knowledge about how that connectivity is provided.
However, even though the Internet is provided ubiquitously, there is still a tremendous need for local networking.
In the same way that companies run their own internal networks today, businesses in the future will be running private clouds. Although it ultimately will happen, we are currently far from fulfilling the entire promise of cloud computing as a new computing model. Nevertheless, there are characteristics that are already becoming evident.
One is that cloud computing now exists in all major computing environments: publicly, selectively, and privately. Amazon is currently the premier provider of public cloud computing; AWS is available to any willing consumer of it. Internet service providers and telecom companies are building cloud environments selectively, i.e. for their existing customer base. Corporations and government agencies with large and varying computing needs are establishing their own private cloud computing environments inside their own firewalls and VPNs. NASA’s Nebula cloud is one such example.
At this early stage of the evolution of this industry, the various initiatives are separate from each other, and it seems rational to distinguish between public and private clouds. And certainly there will be technical differences between a cloud that is designed for anyone willing to consume it and a cloud that is designed for the internal (massive) use by a corporation with a single vision and mission. But over time we will see standardization happening. Customers will look for solutions that can transcend cloud boundaries regardless of where the clouds are sited.
In summary: Private cloud computing software is an essential vehicle for bringing corporate data centers to the cloud era, and over time we are likely to see domination by the clouds which allow easy transfer of computing load across private, hybrid and public clouds.
Cloud Computing: A Promise – Not a Definition
Peter Coffee, Director of Platform Research, salesforce.com
The best way to get into a pointless argument about cloud computing is to think that cloud computing is a technology. It is not. Cloud computing is a set of promises that service providers make to their customers: if a service offering keeps those promises, the label of cloud computing is deserved.

  • Cloud computing is a promise that the limits and the delays of a capital budgeting process will not destroy the momentum of a business process initiative.
  • Cloud computing is a promise that the burdens of irrelevant complexity, the constant updating and patching of a brittle software stack, will be borne by the service provider so the customer can focus scarce resources on the creation of competitive advantage.
  • Enterprise cloud computing, as opposed to the offerings of the consumer Web, is a promise that the needs of the business process – robust security, large-scale capacity, rich customizability – will not be ignored in the pursuit of economies of scale.

Do these promises lead to the development and use of certain key technologies? Of course. Consider these examples:

  • Management by the service provider implies high bandwidth delivered via non-proprietary network protocols: in a word, the Internet.
  • Scalability implies the flexible re-allocation of computing resources to different tasks on short time scales: in a phrase, the use of virtualization.
  • The combination of enterprise flexibility with massive sharing economies requires clear separation of business process customizations from core code base functions such as database maintenance and security model enforcement: in short, the use of a metadata model that can represent and segregate customer-specific content and behavior.

What then of the question of “private clouds”? It’s clear that some of the promises of the cloud can be achieved with either facilities that one owns, or with facilities that are owned and operated by others. For example, an on-premise data center can be remotely managed to receive pushed software updates from an outside source. The question, though, is whether the business model will also keep the cloudy promise of packaging the entire service in a predictable, usage-based pricing scheme – because absent that predictability of cost, this is not cloud computing, but is merely a form of traditional outsourcing.
One cannot look at a collection of servers and software and say, “That’s just a data center” or alternatively, “Aha! That’s a private cloud.” This is precisely as impossible, and for precisely the same reasons, as looking at an automobile and seeing whether it is an owner-driven vehicle (all burdens of maintenance on the owner) or a rented vehicle (usage-based pricing, but still customer-operated) or a chauffered vehicle (usage-based pricing and full service provision to a customer who only worries about where to go). Without a clear statement of the operational responsibilities and the business model and pricing scheme, it’s just hardware. With the right assignment of responsibilities and the right pricing plan, it’s a service.
Cloud computing is a model of service delivery. Remember that, and many arguments simply go away.
M.R. Rangaswami is publisher of SandHill.com .

Copy link
Powered by Social Snap