Cloud Computing: Not Always the Best Solution
Mar 9, 2009 4:00 AM PT
A panel of experts was assembled at The Open Group's Enterprise Cloud Computing Conference in San Diego to examine how cloud computing aligns with enterprise architecture.
The discussion raised the question: What will real enterprises need to do to gain savings and productivity in the coming years to exploit cloud computing resources and methods. In essence, this becomes a discussion about real-world cloud computing.
To gain deeper insights into how IT architects can bring cloud computing benefits to their businesses, I queried panelists Lauren States, vice president in IBM's Software Group; Russ Daniels, vice president and CTO Cloud Services Strategy at HP, and David Linthicum, founder of Blue Mountain Labs.
Listen to the podcast (62:01 minutes).
Here are some excerpts:
David Linthicum: You need to assess your existing architecture. Cloud computing is not going to be a mechanism to fix architecture. It's a mechanism as a solution pattern for architecture. So, you need to do a self-assessment as to what's working, and what's not working within your own enterprise, before you start tossing things outside of the firewall onto the platform in the cloud.
Once you do that, you need to have a good data-level understanding, process-level understanding, and a service-level understanding of the domain. Then, try to figure out exactly which processes, services, information are good candidates for cloud computing.
Not everything is applicable for cloud computing. In fact, 50 percent of the applications that I look at are not good candidates for cloud. You need to consider that in the context of the hype.
Lauren States: The other aspect that's really important is the organizational governance and culture part of it, which is true for anything. It's particularly true for us in IT, because sometimes we see the promise of the technology, but we forget about people.
In clients I've been working with, there have been discussions around, "How does this affect operations? Can we change processes? What about the work flows? Will people accept the changes in their jobs? Will the organization be able to absorb the technology?"
Enterprise architecture is robust enough to combine not only the technology but the business processes, the best practices, and methodologies required to make this further journey to take advantage of what technology has to offer.
Russ Daniels: It's very easy to start with technology and then try to view the technology itself as a solution. It's probably not the best place to start. It's a whole lot more useful if you start with the business concerns. What are you trying to accomplish for the business? Then, select from the various models the best way to meet those kinds of needs.
When you think about the concept of, "I want to be able to get the economies of the cloud -- there is this new model that allows me to deliver compute capacity at much lower cost," we think that it's important to understand where those economics really come from and what underlies them. It's not simply that you can pay for infrastructure on demand, but it has a lot to do with the way the software workload itself is designed.
There's a huge economic value ... if the software can take advantage of horizontal scaling -- if you can add compute capacity easily in a commodity environment to be able to meet demand, and then remove the capacity and use it for another purpose when the demand subsides.
There's a particular class of services, needs for the business, that when you try to address them in the traditional application-centric models, many of those projects are too expensive to start or they tend to be so complex that they fail. Those are the ones where [cloud computing] is particularly worthwhile to consider, "Could I do these more effectively, with a higher value to the business and with better results, if I were to shift to a cloud-based approach, rather than a traditional IT delivery model?"
It's really a question of whether there are things that the business needs that, every time we try to do them in the traditional way, they fail, under deliver, were too slow, or don't satisfy the real business needs. Those are the ones where it's worthwhile taking a look and saying, "What if we were to use cloud to do them?"
Linthicum: Lots of my clients are building what I call rogue clouds. In other words, without any kind of sponsorship from the IT department, they're going out there to Google App Engine. They're building these huge Python applications and deploying them as a mechanism to solve some kind of a tactical business need that they have.
Well, they didn't factor in maintenance, and right now, they're going back to the IT group asking for forgiveness and trying to incorporate that application into the infrastructure. Of course, they don't do Python in IT. They have security issues around all kinds of things, and the application ends up going away. All that effort was for naught.
You need to work with your corporate infrastructure and you need to work under the domain of corporate governance. You need to understand the common policy and the common strategy that the corporation has and adhere to it. That's how you move to cloud computing.
States: The ROI that we've done so far for one of our internal clouds, which is our technology adoption program, providing compute resources and services to our technical community so that they can innovate, has actually had unbelievable ROI -- 83 percent reduction in cost and less than 90-day payback.
We're now calibrating this with other clients who are typically starting with their application test and development workloads, which are good environments because there is a lot of efficiency to be had there. They can experiment with elasticity of capacity, and it's not production, so it doesn't carry the same risk.
Daniels: Our view is that the real benefits, the real significant cost savings that can be gained. If you simply apply virtualization and automation technologies, you can get a significant reduction of cost. Again, self-service delivery can have a huge internal impact. But a much larger savings can be done if you can restructure the software itself so that it can be delivered and amortized across a much larger user base.
There is a class of workloads where you can see orders-of-magnitudes decreases in cost, but it requires competencies, and first requires the ownership of the intellectual property. If you depend upon some third-party for the capability, then you can't get those benefits until that third-party goes through the work to realize it for you.
Very simply, the cloud represents new design opportunities, and the reason that enterprise architecture is so fundamental to the success of enterprises is the role that design plays in the success of the enterprise.
The cloud adds a new expressiveness, but imagining that the technology just makes it all better is silly. You really have to think about, what are the problems you're trying to solve, where a design approach exploiting the cloud generates real benefits.
Dana Gardner is president and principal analyst at Interarbor Solutions, which tracks trends, delivers forecasts and interprets the competitive landscape of enterprise applications and software infrastructure markets for clients. He also produces BriefingsDirect sponsored podcasts. Disclosure: The Open Group sponsored this podcast.