Tuesday, October 22, 2013

Open Source Not a Panacea for Cloud Infrastructure Decisions

Cloud IaaS Open Source
When it comes to cloud computing, do open systems win out over proprietary standards? My view is, perhaps in theory, but cloud computing--specifically public cloud infrastructure--has bigger problems right now than whether it's built on open source. Furthermore, open source cloud infrastructure providers have obstacles to overcome. 

I'm participating in an online video debate on October 29, hosted by IBM's Smarter Computing program, on "the pros and cons of open computing when it comes to cloud, big data, and software defined environments." This post outlines part of my viewpoint on this subject.

What's Not to Like about Open Source?

One of the problem in debating "open source" is that it is difficult to argue against the word "open" as a concept. For example, we all like to think of ourselves as open-minded, not close-minded. We admire top executives who have an open-door policy--have you ever heard of a manager with a "closed door policy?" In home-buying, sellers like to point out the open floor plan. Who ever advertised a house as having a "closed" floor plan?

So also, in computing, open just sounds better. Moreover, when it comes to cloud infrastructure, open source projects such as OpenStack and CloudStack have admirable goals, such as the ease of porting computing workloads from one cloud provider to another, promoting competition, and escaping the dreaded vendor lock-in.

The Larger Issue: Adoption

But, to me, it is premature to debate about whether open source cloud infrastructure is better. The larger issue today is the small percentage of corporate IT organizations that embrace public cloud infrastructure at all. In our Technology Trends survey at Computer Economics last year, we found that less than 10% of IT organizations worldwide have any use or plans to use public cloud infrastructure. Moreover, of these, only half claim use it, or intend to use it, for production systems.

If they are not using public cloud for production systems, then what are they using it for? Our survey found interest in public cloud for software development and testing, disaster recovery capabilities (such as backup and recovery), or for archiving older data.

In addition, I question some of those production uses of IaaS. Discussions with associates who advise data center managers confirm my suspicions. One associate, who works a lot in the entertainment industry, pointed out that one popular use of cloud infrastructure is in rendering animated film. In this case, animators require enormous amounts of computing power and storage to render even a few minutes of animation. As it turns out, cloud infrastructure is perfect for such a use, as it frees the IT organization from having to maintain those high levels of computing resources, which are only used sporadically. Furthermore, the risk is low. If the cloud provider goes down in the middle of a rendering job, the animator can simply resubmit the job. Nothing is lost.

But when it comes to production systems, such as accounting systems or royalty processing, these same entertainment industry decision-makers shun cloud infrastructure. It is not that they want to keep such systems on-premises, as witnessed by the fact that they have been outsourcing their data centers to managed services providers for years. As my associate remarked, "CIOs don't want to be in the data center business any more." But, rightly or wrongly, they are cautious about entrusting production systems to a cloud infrastructure.

Open Source Not a Panacea

Although the goals of OpenStack and other open source cloud projects are admirable, they may be a solution in search of a problem.
  • Specifically, migrating workloads between competing cloud providers may not be as big a deal as open source proponents claim. Customer demands are already forcing competing cloud providers to recognize and support each other's APIs. For example, some members of the OpenStack community are urging support for Amazon's APIs.  If OpenStack fully goes this route, application systems written for Amazon's cloud will be able to be deployed on an OpenStack cloud without a lot of migration effort. Even VMware--the vendor with the largest stake in so-called private clouds--supports Amazon APIs and is also a contributor to OpenStack. Therefore, as far as I can tell, portability is not a major issue.
  • Second, so far, it does not seem as if proprietary cloud providers are using their proprietary standards in order to extract higher fees from customers. Quite to the contrary, cloud infrastructure is a very competitive market. Whatever concerns IT decision makers have about public cloud infrastructure, one thing they cannot complain about is its cost. Leading cloud providers are not raising prices--rather, they are cutting prices, in some cases many times a year. IT decision makers are not holding on to their on-premises systems because they are concerned about the cost of public cloud--they are focused on risk. This was also a key finding in our Technology Trends survey.
If a cloud provider wants to overcome enterprise IT buyer concerns, it should focus on reliability, security, privacy, and offer a well-staffed support group. Many of the OpenStack providers are doing exactly that. It may well be that OpenStack providers, such as IBM, H-P, Dell, Rackspace and others, will be successful because of their value-added services, not because they embraced an open source infrastructure.

Incumbent Infrastructure Providers Have an Edge

Furthermore, proponents of open source cloud infrastructure may be underestimating the advantage that on-premises infrastructure providers have in moving their customers to the public cloud. Although, as discussed above, IT leaders have concerns about moving production workloads to the public cloud, one thing that does appeal to some of them is the ability to move seamlessly from on-premises system instances to cloud instances.

This is the so-called hybrid cloud infrastructure. CIOs may adopt a hybrid cloud strategy in order to move non-critical workloads out of the data center, freeing up system resources (e.g. the animation rendering application discussed above), or to "burst" to the cloud during period of high demand for system resources (e.g. during a major advertising campaign that strains an in-house e-commerce system).

Now, which provider has the advantage in helping IT organizations set up hybrid cloud capabilities? The provider that is already serving the on-premises data center (Microsoft, VMware, or Oracle, for example) or the one that would like the data center to replatform its on-premises systems to match the infrastructure of the provider's cloud infrastructure (e.g. OpenStack, CloudStack)?

The answer is obvious, which is why Microsoft, VMware, and Oracle are all providing public cloud services that require very little change to the customer's on-premises infrastructure. Unless an IT organization is building a data center from scratch, it is unlikely to want to standardize its internal infrastructure on a completely new technology--open source or otherwise.

Advocating for Cloud and Open Source

Nothing I've written here should be taken as an argument against cloud computing or open source. I've been blogging on these subjects since 2002 and consider myself as an advocate of both. In my view, one day nearly all systems will be delivered via cloud computing, and open source software has proven itself to be a viable business model for a variety of software categories, especially for lower levels in the technology stack. But in the case of public cloud infrastructure, I don't see open source cloud projects as dominating the market any time soon.

Update: The video of my IBM debate is now online. You can watch it by clicking the image below.

http://www.spreecast.com/events/whats-next-for-it-open-source-debate

Related Posts

The Inexorable Dominance of Cloud Computing
Cutting Through the Fog of Cloud Computing Definitions

Photo Credit: Flickr, followtheseinstructions

2 comments:

clive boulton said...

Another way of inspection this is comparison of traditional slow growth on premise backend ERP and high-growth front end online eCommerce. At some point all the front end business needs routing to ERP--here we hit impedance mismatch between promise and delivery (or expectation and fulfillment).

Logically ERP could move to the cloud to make communication easier than trying to reliably connect online with onprem business.

If we flatten more bills-of-material into "Dell" configurable / deliverable MRP. Then by adopting Cloud we can ramp up the growth engine and economic profitability.

Alas, I can't understand why SMEs are so reluctant move their Sales online and take advantage of all the open source cross vendor collaborative efforts.


Behold, Frank?

Frank Scavo said...

Clive, my observation is that with most companies, especially SMEs, there is much attitude of "if it ain't broke, don't fix it." Unless a system is really broken, or there is some other driving event, companies are loath to swap out existing systems. Once they do encounter a driving event, however (e.g. a merger or acquisition), everything is open, and that is the point at which cloud systems become very attractive.