Saturday, October 31, 2009

The inexorable dominance of cloud computing

Cloud computing is not just one more way to deploy information systems. It represents a total shift in how IT resources are delivered and ultimately will replace most of not all internally-maintained IT infrastructure.

At least that's the view of Nicholas Carr, who gave a talk at a one-day conference on cloud computing organized last week in London by Google. If you've read Carr's work in the past, his presentation will be familiar. One main point: the on-premise deployment of systems such as Oracle and SAP today are analogous to the on-premise factory power-plants of the 19th century--ultimately replaced by public electric utilities. So, it will be with utility computing.

Carr also gets a little bit into IT budget ratios, which we track closely at Computer Economics. His analysis is spot on and is a strong argument for why cloud computing ultimately will prevail over on-premise systems.

Near the end of the presentation, Carr presented five models for adoption of cloud computing:
  1. Internal clouds: large organizations take advantage of cloud computing technologies by moving their own large IT infrastructures to a cloud computing model.

  2. Cloud as supplement: organizations retain their on-premise systems but use cloud computing to deploy new IT capabilities.

  3. Cloud as replacement: organizations forgoing their own IT infrastructure altogether and going with cloud computing for everything. So far this is appealing, naturally, to smaller businesses.

  4. Cloud as democratizer: cloud computing allowing individuals to have their "own data centers," leading to an explosion in innovation.

  5. Cloud as revolution: cloud computing reducing the cost and increasing the accessibility of data processing, leading to new ways of embedding IT in new products and services.
You can view Carr's 30 minute talk here:



What about the major on-premise software providers, such as SAP and Oracle? Can they make the transition to cloud computing? Although both SAP and Oracle have cloud computing initiatives, such as SAP's Business ByDesign and Oracle's On-Demand CRM, I'm not hopeful. They have too much invested in, and receive too much of their margin, from their legacy products.

Consistent with this view, at the end of his presentation, Carr looks at cloud computing as a disruptive technology, a la Clayton Christensen. I was glad to hear that, as I've long felt that cloud computing strongly qualifies as a disruptive technology that does not give hope to the current market leaders in enterprise software. But that's the subject for another post.

From time to time I bring up issues that buyers should be aware of in evaluating SaaS providers--for example, business continuity concerns. But in the long run, I'm convinced these issues will be worked through. The transition will take some time, but the economics are too strong for cloud computing not to prevail. Therefore, even today, buyers should consider every IT decision in light of the options available in the cloud.

Update, Nov 1. Be sure to read the first comment on this post, from former SAP executive Nenshad Bardoliwalla, who essentially confirms my point.

Related posts
Salesforce.com: more than an itty-bitty application
NetSuite a viable alternative for SAP customers?
Cloud computing: can Microsoft turn from servers to services?
IT departments face extinction
The end of corporate computing
Computer Economics: The Business Case for Software as a Service

2 comments:

Nenshad Bardoliwalla said...

Hi Frank,

Jon Reed and I did a podcast on the topic of SAP and cloud computing that might be worth your time to listen to. It's located here: http://bit.ly/2KiNl6 . You can also read a transcript of the conversation here: http://bit.ly/2070Ln

From the transcript:

There is no doubt that there is an aura of inevitablity about SaaS and cloud computing, because the value proposition is significantly better than on premise software. With on premise, you assume all the risk, assuming responsibility for installation, training, and patches. With SaaS, the vendor has to handle security, skills needs, patches, installation, compatibility issues.

Removing all the hype, in many cases, the cloud model is better for the customer, so cloud computing will prevail over time and there is a very large shift underway. Example: it's almost impossible to get a venture capitalist to fund an on premise software solution right now. SaaS solutions have a much better chance of being funded.

But that being said, we're still in the infancy of the maturity of cloud-based solutions. Example: Salesforce.com has tens of thousands of customers on one Oracle replica database. SAP has customers with single instances of SAP ERP whose Oracle database is larger than all of Salesforce.com. The idea that with today's technology and cloud computing that a vendor can come in and replace that doesn't make sense.

Years from now, cloud systems will be so powerful that a real alternative to SAP and Oracle on a larger scale. People forget the scale of some of the current deployments of SAP and Oracle, and they forget the business processes that must be run on premise versus the cloud. Example: APO.

There are customers running APO, a statistically rich product, which require one terabyte of main memory to run APO correctly. This is the amount of memory needed to run these scenarios. If you needed to run a processes that required that much memory, it would probably take down all the other tenants in a multi-tenant architecture. Pioneers like Workday and NetSuite, going for the full ERP replacement opportunity, are not ready yet to replace the major big ERP installations. But the change is underway, and more and more customer spend will move to the cloud.

Best Regards,

Nenshad

Matthew King said...

Some projects I've seen, use large numbers of CPUs for parallel processing but only utilize them for a few hours a day. Being able to share this hardware with others in different time zones would improve CPU utilization.

Same could be said for sharing main memory, but to a lesser extent because some main memory in planning systems is occupied even when CPUs are idle.

If a group of large customers spread across multiple time zones moved to a multi-tenant model, the hardware savings alone could be huge. The application architecture would need to support such an approach, of course. And currently available hardware units would limit the number of sharing tenants, but this limitation would reduce over time as ever larger capacity units comes to market.

Also, large multinationals with dedicated (and centrally located) systems for each region would also benefit from a multi tenant architecture, allowing them to better utilize their own in house (on premise) hardware.