Tuesday, September 28, 2010

Automated Governance: The Lynchpin for Success or Failure of Cloud Computing

Transcript of a sponsored podcast on cloud computing and the necessity for automated and pervasive governance across a services lifecycle.

Listen to the podcast. Find it on iTunes/iPod and Podcast.com. Download the transcript. Get a copy of Glitch: The Hidden Impact of Faulty Software. Learn more about governance risks. Sponsor: WebLayers.

Dana Gardner: Hi, this is Dana Gardner, Principal Analyst at Interarbor Solutions, and you're listening to BriefingsDirect.

Thanks for joining this sponsored podcast discussion on why governance is so important in the budding era of cloud computing. As cloud-delivered services become the coin of the productivity realm, how those services are managed as they are developed, deployed, and used -- across a services lifecycle -- increasingly determines their true value.

Management and governance are the arbiters of success or failure when we look across a services ecosystem and the full lifecycle of those applications. And yet, governance is still too often fractured, poorly extended across the development-and-deployment continuum, and often not able to satisfy the new complexity inherent in cloud models.

One key testbed for defining the role and requirements for cloud governance is in applications development, which due to the popularity of platform as a service (PaaS) is already largely a services ecosystem.

Often times, development teams are scattered globally, contractors can come and go, testing is provided as a service -- all while the chasm between development and deployment shrinks and iterations of deployments are hastening.

Here we’ll discuss the needs and potential for solutions around governance in the cloud era using the development and deployment environment as a bellwether for future service environments.

Here to help us explain why visibility across services creation and deployment is essential -- and how governance can be effectively baked into complex ecosystems -- we're joined by Jeff Papows, President and CEO of WebLayers and the author of Glitch: The Hidden Impact of Faulty Software. Welcome back to BriefingsDirect, Jeff.

Jeff Papows: Dana, thanks for having me on again.

Gardner: And, we're also here with John McDonald, CEO of CloudOne Corp. Welcome to the show, John.

John McDonald: Dana, hi. Thanks.

Gardner: Let’s start off, as it often the case with these cloud discussions, by defining "cloud" for our purposes, what we're going to focus on today, and I think that has lots to do with PaaS. So, let’s start with you John McDonald. Tell us what you think of when people mention cloud, particularly in development and deployment strategies in this notion of PaaS?

The role of confusion

McDonald: There is a ton of confusion about this right now and to be honest with you, for a lot of companies this confusion serves them in what they are trying to do.

To try to clarify it for everybody, cloud computing is really quite simple to understand. It’s all about getting access to hardware on-demand. This is hardware that I might use for any number of purposes -- for storing data, providing tools, and hosting an application.

There are a lot of companies out there that have done hosting in the past, application hosting or whatever, who are now morphing into cloud-computing companies. Some of them are actually even using cloud-computing technologies to do it, even though they just named themselves that.

Cloud, from a technology perspective, is more about some very sophisticated tools that are used to virtualize the workloads and the data and move them live from one bank of servers to another and from one whole data center to another, without the user really being aware of it. But, fundamentally, cloud computing is about getting access to a data center that’s my data center on-demand.

It’s frequently confused with another concept called software as a service (SaaS). SaaS is about getting access to software on-demand. So as cloud is to hardware, SaaS is to software. Frequently, these concepts are used together, so that when you do that you have an environment that scales up and down dynamically as your needs change up and down.

Sometimes that’s labeled Platform as a Service. In other words, I'm providing an entire platform or a work bench of tools on-demand. There are two concepts together, sometimes it’s called infrastructure as a service (IaaS), when what it is that I am providing is more of an infrastructural set of tools.

Fundamentally, the easiest way to remember it is that cloud is to hardware as SaaS is to software. Basically, for CloudOne, we're providing IBM Rational Development tools both through cloud computing and SaaS. Right now, we're the only people doing that. So, it’s unique and frankly pretty fun.

Gardner: Jeff Papows, why do you think application development has become such a great demonstration of what cloud computing can do? Why is there such a good fit between cloud, as far as John just defined it, and application development?

Papows: John’s explanation was both accurate and important, because there's a habitual capacity in our industry, as both of you have recognized, for people to get confused or hung up on vocabulary and on the most recent flavor of acronym headaches.

If you think about a lot of what John said, and a lot of about what’s going on in cloud computing it’s not a particularly new thing. What we used to think of was hosting or outsourcing. Then, you saw vertical instantiations of it around particular competencies like payroll. Companies like ADP were basically clouds with distinctive vertical expertise and processing payroll and doing tax reporting.

Mobile world

What’s happening now is the world is becoming more mobile, as 20 percent of our IT capacity is focused on new application development every year as opposed to maintaining what we have.

We have to get more creative and more distributed about the talent that contributes to those critical application development and projects. That’s why you begin to see, as John started to describe it, a razors and razor blade taxonomy, where it’s one thing to virtualize the hardware environment and some of the baseline topology and infrastructure, but then you begin to add layers of functionality.

Rational Team Concert (RTC) is one good case in point, as John pointed out too, but design time governance is the next logical thing in that continuum, so that all of the inherent risk mitigation associated with governance and then IT contacts can be applied to application development in a hybrid model that’s both geographically and organizationally distributed.

Gardner: John McDonald, you mentioned the fact that cloud fits in where workloads are unpredictable. With application development that’s certainly the case. It’s not just the constant hum of production, but really more fits and starts. Tell me, from your perspective, why cloud works so well to support application development across its continuum right up and into deployment?

McDonald: Yeah, that is the case. There's a myth that development is something that we ought to be tooling up for, like providing power to a building or water service. In reality, that’s not how it works at all.

The money that you save by doing that is the reason you can open any trade magazine and the first seven pages are all going to be about cloud.

There are people who come and go with different roles throughout the development process. The front-end business analysts play a big role in gathering requirements. Then, quite often, architects take over and design the application software or whatever we are building from those requirements. Then, the people doing the coding, developers, take over. That rolls into testing and that rolls into deployment. And, as this lifecycle moves through, these roles wax and wane.

But the traditional model of getting development tools doesn’t really work that way at all. You usually buy all of the tools that you will ever need up front, usually with a large purchase, put them on servers, and let them sit there, until the people who are going to use them and log in and use them. But, while they are sitting there, taking up space and your capital expense budget, and not being used, that’s waste.

This cloud model allows you to spin up and spin down the appropriate amount of software and hardware to support the realities of the software development lifecycle. The money that you save by doing that is the reason you can open any trade magazine and the first seven pages are all going to be about cloud.

It's allowing customers of CloudOne and IBM Rational to use that money in new, creative, interesting ways to provide tools they couldn't afford before, to start pilots of different, more sophisticated technologies that they wouldn't have been able to gather the resources to do before. So, it's not only a cost-savings statement, it's also ease of use, ease of start-up, and an ability to get more for your dollar from the development process. That's a pretty cool thing all the way around.

Gardner: So the good news is about that agility, that flexibility and adaptability toward a workflow of some sort across a development process. The bad news is that these things can spin out of control and that there is not a common thread or a fabric around them -- especially if you're doing source in your cloud’s hybrid models or multiple cloud or multiple sources of the platform or tools or testing.

Back to you, Jeff Papows. What do we do in terms of defining the problem set? What's the problem that governance is going to help solve?

Economic perturbation

Papows: John describes some of the economic realities, as well as the pragmatic realities of agile development, which I agree is not linear. It's a set of perturbations that, as John said, wax and wane depending on where you are in a particular development cycle, in which organizations your skills are being, are being amassed. That's as it should be, and it's nature’s law. In any event, you're not going to change it.

When you try to add some linear structure and predictability to those hybrid models, as you both have been discussing, the constant that can provide some order and some efficiency is not purely technology-based. It's not just the virtualization, the added machine capacity, or even the middleware to include companies like WebLayers or tools like Rational. It's the process that goes along with it. One of the really important things about design-time governance is the review process.

In a highly distributed, hybrid, agile application-development model -- where you may have business analysts in Akron, Ohio, architects in Connecticut, coders in Singapore, and outsourced QA in India -- the one constant taxonomy is the ability to submit and review and deal with some logical order and structure to the workflow that makes that collaborative continuum more predictable and more logical, irrespective of all of the moving parts, both digital and human, and the fabric that we're talking about here.

Governance is a big part of the technology toolset that institutionalizes that review process and adds that order to what otherwise can quickly become a bit chaotic, depending on where you are in the perturbations that John describes.

McDonald: This is a really good point that you're making, Jeff. The challenge of tools in the old days was that they were largely created during a time where all the people and the development project were sitting on the same floor with each other in a bunch of cubes in offices.

The cloud allows us to create a dedicated new data center that sits on the Internet and is accessible to all, wherever they are, and in whatever time zone they are working, and whatever relationship they have to my company.

As the challenges of development have caused companies to look at outsourcing and off-shoring, but even more simplistically the merger of my bank and your bank, then we have groups of developers in two different cities, or we bought a packaged application, and the best skill to help us integrate it is actually from a third-party partner which is in a completely different city or country. Those tools have shown their weaknesses, even in just getting your hands on them.

How do I punch a hole through the firewall to give you a way to check in your code problems? The cloud allows us to create a dedicated new data center that sits on the Internet and is accessible to all, wherever they are, and in whatever time zone they are working, and whatever relationship they have to my company.

That frees things up to be collaborative across company boundaries. But with that freedom comes a great challenge in unifying a process across all of those different people, and getting a collaborative engine to work across all those people.

Papows: That’s a great point John. I was with the CIO of a major New York bank about two weeks ago. Like so many CIOs in this financial services sector post-2008 they are in the midst of clamming together two very large complex inherently different back-office systems. Then, on a magical date, somehow they're supposed to intersect without the digital version of Pearl Harbor. That’s not a reasonable request, but these are not reasonable times.

Complexity curve

Without the ability to create these ad hoc environments, not just organizationally or geographically, but perhaps separate production from testing and development -- and without the ability to automate a good part of the tooling associated with reviewing these massive, mountains of legacy code bases before you magically intersect these things and put them together -- there's not a prayer that carbon-based, biped life forms are going to pull that off without a far more automated approach to that kind of a problem. It’s reached a point in the complexity curve, where you just can’t throw enough bodies at it.

McDonald: That’s right. It’s almost a requirement to keep the wheels on the bus and to have some degree of ability to manage the process in the compliance with regulations and the information about how decisions were made in such distributed ways that they are traceable and reviewable. It’s really not possible to achieve such a distributed development environment without that governance guidance.

Gardner: One of the interesting things that I have noticed in talking about cloud for the past several years is the realization fairly early on that the owner of the application or service -- and not the provider -- are the ones who are inherently and ultimately responsible for the governance. They are also responsible to the end user in in terms of their performance expectations. So, given that reality, who is responsible for governance and where should it begin and end? Where does it intersect with ownership within these ecosystems?

Papows: When I say "governance," I'm not talking about the Sarbanes-Oxley corporate governance context. I am talking about it specifically as it relates to IT. That is a function of the C-level executives, meaning it’s a partnership between the CIO and the CEO. This is not something that happens at the level of architect, the program, or this digital professional that’s in the trenches. There is an aspect of this, Dana, that we have to wake up and get environmentally much more honest with one another about.

We're dealing with some challenges for the first time that require out-of-the-box thinking. I talk about this in "Glitch." We have reached a point where there a trillion connected devices on the Internet as the February of this year. There are a billion embedded transistors for every human being on the planet.

We have reached a point where there a trillion connected devices on the Internet as the February of this year. There are a billion embedded transistors for every human being on the planet.

For the first time, we're seeing a drought in available computer scientists graduating from colleges and universities. The other side of the dot-com implosion was that the vocation became somewhat less attractive to people.

Moreover, 70 percent of the transaction-processing systems that we're dependent upon in the world economy today run on the things like mainframes and they are written in languages like COBOL. Although there are some very valiant efforts being made by IBM in about 600 universities, we're going to see more of that human capital retire, reach the end of their time with us, and die off in terms of the workforce. Yet, all of that inherent complexity is, at the same time, being exacerbated by all of these mergers and acquisitions.

Put all of those things together and, if it weren’t for companies like CloudOne that are creating these ecosystems and distributed environments that allow people to deal with the 20 percent of that new application development in unique and new ways vis-à-vis the cloud, for the first time in the history of our industry, as computer scientists, we're on the verge of tremendous challenges. That’s why I say it’s a partnership between C-level executives, because these are not traditional times.

Gardner: John McDonald, where do you see the notion of baking-in governance taking place. Clearly, the incentive, the direction, and the vision needs to come from on-high. But how do you embed governance into a development workflow, for example?

Everything has to disappear

McDonald: My view is that it absolutely has to be so incipiently based that everything that you are doing has to disappear. Here’s what I mean by that. Developers view themselves quite often as artists. They may not articulate it that way, but they often see themselves as artists and their palette is code.

As such, they immediately rankle at any notion that, as artists, they should be governed. Yet, as we’ve already established, that guidance for them around the processes, methods, regulations, and so on is absolutely critical for success, really in any size organization, but beyond the pale in a distributed development environment. So, how do you deal with that issue?

Well, you embed it into their entire environment from the very first stage. In most companies, this is trying to decide what projects we should undertake, which in lot of companies is a mainly over-glorified email argument.

It goes right on through to the requirements gathering around those projects that we have decided to undertake to the project plans that are put around those projects, to the architecture, the design, the coding, the testing, the build, and the deployment.

It all has to be embedded at every step of that way, gently nudging, and sometimes shuttling all these players back into the right line, when it comes to ensuring that the result of their effort is compliant with whatever it is that I needed to be compliant to.

In short, Dana, you’ve got to make it be a part of and embedded into every stage of the development process, so that it largely disappears, and becomes something that becomes such a natural extension of the tool so that you don’t have anyone along the way realizing that they are being governed.

Unless you automate it, unless you provide the right stack of tools and codify the best practices and libraries that can be reusable, it simply won’t happen.

Papows: John is exactly right, Dana. It’s got to be automated. You’re not going to do something as ubiquitously as John is describing in a manually intensive non-electronic process. It will fundamentally break down.

Everybody intellectually buys into governance, but nobody individually wants to be governed. Unless you automate it, unless you provide the right stack of tools and codify the best practices and libraries that can be reusable, it simply won’t happen. People are people, and without the automation to make it natural, unnatural things get applied some percentage of the time, and governance can’t work that way.

Gardner: Let’s look at an example vis-à-vis CloudOne. John, tell me a little bit about how you do this. Now that we’ve made a determination to this is the right approach, I’m assuming you use WebLayers to do this. Tell me a little bit about CloudOne as an example of how this can work.

McDonald: When we first began this company, all those many months ago, we knew that this is going to be incredibly important.

WebLayers was the very first partner that we reached out to say, "Can you go down this journey with us together, as we begin developing these workbenches, these integrated toolsets, and delivering them through the cloud on-demand?" We already know and see that embedding governance in every layer is something we have to be able to do out of the gate.

The team at WebLayers was phenomenal in responding to that request and we were able to take several based instances of various Rational tools, embed into them WebLayers technology, and based on how the cloud works, archive those, put them up in our library to be able to be pulled down off-the-shelf, cloned, and made an instance of for the various customers that we have coming to our pipeline who want to experience this technology in what we are doing.

So, right from the start, Dana, we put that into what we are doing, so that when customers experience CloudOne’s technology either in pilot or in production they never know that it’s not theirs. CloudOne Team Concert is a better Rational Team Concert, because it has WebLayers embedded into it, than simply buying Team Concert and doing it on your own.

Embedded automation

At this point, we have approaching a hundred customers who have, in one shape or form, used or touched some WebLayers technology in the course of a pilot. We frankly see a very healthy group of customers, as we go into the fall of this year, who we believe are going to become customers of that technology simply because they have been able to experience that embedded automation, almost disappearing into the background kinds of guidance for governance is what Jeff has been talking about. So, it’s been really a great journey so far, and I can only see it getting better.

Gardner: I know it’s hard to quantify results when you are preventing something bad from happening, but are there any metrics of success? Can you point to the embedded governance and say that got us to "blank" or paid off in some manner or another?

Papows: Unfortunately, the best examples of those tend to come from the places where governance is not. You’ve read about or heard about or experienced first hand the disasters that can happen in production environments, where you have some market-facing application, where service is lost, where there is even brand damage or economic consequences.

We’ve seen ad hoc development. As an example, a year ago at a major European investment banking firm where a CEO did what CEOs frequently do, and demanded that everyone complete an online workflow for everyone’s annual review process.

This particular CEO went further and said that if it wasn't done by year-end, any manager who hasn’t completed this for all of his or her constituents wouldn’t be eligible for the year-end bonus. Somebody very quickly cobbled together some HR workflow, unbeknownst to anyone.

It’s seven times more expensive to fix an application service after it’s deployed than it is in design.

There was a sense of urgency. It relied on a single database thread that was part of a production system that was not reinforced. When everybody collided at once to coalesce to the demands that this particular executive was articulating, it brought down the trading floor.

In the four hours that that system was lost, as Murphy’s law would frequently have it, there was about an 11 percent market accretion. The cost of that particular institution for the difference in the trading value for the hours that they were out of business was about $24 million.

There are instances like that, which become almost water-cooler legend, where you can quantify fantastic ROI in reverse. There is a new concept -- and John is probably starting to get exposed to this -- called "technical debt" -- I think one of your blogs touched on this earlier.

We're beginning to quantify the opportunity cost and the human cost in terms of mandates of IT time for things that are not governed or not adhered, so that as you catalogue the number of programs, files, WSDLs, objects, and stuff that don’t meet the acid test, you get a sense for the number of days, and as a consequence the dollars, of technical debt that you are amassing.

There was a great article -- I can’t remember who published it -- that said that it’s seven times more expensive to fix an application service after it’s deployed than it is in design. God knows whether that’s got any decimal-point accuracy, but it’s certainly directionally correct. We are going to provide some dashboard reporting in some objects in our management dashboard series.

As we look toward the end of the year, that will give you some widgets and dials, and we’ll begin to quantify the cost of the things that we find that don’t adhere to the libraries that people like John are building into their infrastructure. While a lot of things in information technology in the last couple of decades have been largely subjective, we're going to get to the point where we are going to start quantify these things fairly precisely.

Gardner: What about you, John McDonald? Do you have any sense of the paybacks, the metrics of success when governance is done properly in your neck of the woods?

Signboards of success

McDonald: I have to agree with Jeff. The biggest signboards of success here are when things go badly. The avoidance of things going badly is unfortunately very difficult to measure. That is something that everyone who attempts to do a cloud-delivered development environment and does the right thing by embedding in it the right governance guidance should know coming out of the gate. The best thing that’s going to happen is you are not going to have a catastrophe.

That said, one of the neat things about having a common workbench, and having the kinds of reporting in metrics that it can measure, meaning the IBM Jazz, along with the WebLayers technology, is that I can get a very detailed view of what’s going on in my software factory at every turn of the crank and where things are coming off the rails a little bit.

I equate this in some ways to a car production factory, where there are many moving parts, lots of robot arms, and people lifting plate glass into place and screwing in bolts and that sort of thing. Everything may look great in my factory, but at the end of the factory, I consistently see the door handle is off by three inches. I can’t release those cars to my dealership network with bad door handles, so I know that I've got a problem, but I can very quickly see where the problem is. That’s how most companies right now deal with governance issues. They wait until the very end of it, as it’s about ready to be shipped to the dealer, and then they notice the door handle is off.

It may be great to go back and know where to fix the door handle, but wouldn’t it be nice to know, before that car went to the rest of the line, that we had a problem with the machine in the door handle’s section. That’s what the kinds of metrics and measurement and responsiveness that this offers allows you to do -- fix the door handle, before it gets any farther down the line, so you never get to that catastrophe where the engine falls out of the bottom.

You don’t even get to the small issues where the door handle is off. You nip them in the bud. Doing that live every day with the visibility into the reports and the metrics around governance is really the magic here, so that you never have that issue of a catastrophe, where you have to hold up and say, "Well, we’ll do better next time."

There's an age-old expression that you're so close to the forest you can't see the trees. Well, I think in the IT business we’re sometime so deeply embedded in the bark we can't see anything.

Gardner: Let's take a look to the future. Clearly, you'll find few people to argue with the fact that software is becoming more important to more companies. And, the cloud is becoming a new way -- at least new in terms of how people conceive of it -- of acquiring software and delivering services.

So this is all going to get worse. We're going to have more companies that see a strategic imperative around software and development and more opportunity for ecosystems and services. Then, of course, we’ve got explosion of data and mobile devices. Let me go first to you, Jeff.

I suppose I already know the answer, but is this important to do now? It's just going to get worse, but doesn't this also cut across and beyond where we go with development and into so many other areas of business? It seems, as I said in the setup, that you guys are the bellwether of how this can become more prevalent across more aspects of business in general.

Papows: You're right. Here is the reality, and it’s interesting sometimes. There's an age-old expression that you're so close to the forest you can't see the trees. Well, I think in the IT business we’re sometime so deeply embedded in the bark we can't see anything.

We've been developing, expanding, deploying, and reinventing on a massive scale so rapidly for the last 30 years that we've reached a breaking point where, as I said earlier, between the complexity curves, between the lack of elasticity and human capital, between the explosion and the amount of mobile computing devices and their propensity for accessing all of this backend infrastructure and applications, where something fundamentally has to change. It's a problem on a scale that can't be overwhelmed by simply throwing more bodies at it.

Creative solutions

Secondly, in the current economy, very few CIOs have elastic budgets. We have to do as an industry what we've done from the very beginning, which is to automate, innovate, and find creative solutions to combat the convergence of all of those digital elements to what would otherwise be a perfect storm.

That, in fact, is where companies like CloudOne are able to expand and leap productivity equations for companies in certain segments of the market. That's where automation, whether it's Rational, WebLayers, or another piece of technology, has got to be part of the recipe of getting off this limb before we saw it off behind us.

The IT business has become such a critical part of our economy. Put the word "glitch" in your Google Alerts bar and see how many times a day you find out about customers that are locked out of ATM networks, manufacturing flaws, technology disasters in the Gulf, nuclear power plants in Houston, or people being killed by over-radiation because of software bugs in medical equipment. It's reaching epidemic proportions, and the proof-point is that you see it in the daily broadcast news cycles now.

There there is simply no barrier for anyone to give this a try.

So SaaS, cloud computing, automated governance, forms of artificial intelligence, Rational tooling, consistent workbench methodologies, all of these things are the instruments of getting ourselves out of the corner that we have otherwise painted ourselves in.

I don't want to seem like an alarmist or try to paint too big a storm cloud on the horizon, but this is simply not something that's going to happen or be resolved in a business-as-usual usual fashion.

Gardner: Okay, so the stakes are high, they are getting higher. Back to you for the final word, John McDonald. What do you recommend for people who need to get started or are thinking of getting more involved with governance part-and-parcel with their activities?

McDonald: That's one of the coolest things of all about this whole model, in my mind. There there is simply no barrier for anyone to give this a try. In the old model, if you wanted to give the technology a try, you had better start with your calculator. And you had better get the names and addresses of your board of directors, because you're going there eventually to get the capital approval and so on to even get a pilot project started in many cases with some of these very sophisticated tools.

This is just not the case anymore. With the CloudOne environment you can sign on this afternoon with a web-based form to get a instance of let's say, Team Concert set up for you with WebLayers technology embedded in it, in about 20 minutes from when you push "submit," and it's absolutely free for the first model. From there, you grow only as you need them, user-by-user. It's really quite simple to give this concept a try and it's really very easy.

If you have any inclination at all to see what it is that Jeff and I are telling you, give it a whirl, because it's very simple.

Gardner: Okay, we'll have to leave it there. We've been discussing the needs and potential for solutions of governance in the cloud era, and using managed development and deployment environments as a bellwether for future cloud and service and IT use.

I want to thank our guests, we have been talking with Jeff Papows, President and CEO of WebLayers as well as the author of Glitch: The Hidden Impact of Faulty Software. Thanks so much, Jeff.

Papows: Thank you, Dana, and thank you, John.

Gardner: Yes, we've been joined here also by John McDonald, the CEO of CloudOne Corp.

Gardner: I'm Dana Gardner, Principal Analyst at Interarbor Solutions. You've been listening to a sponsored BriefingsDirect podcast. Thanks for listening, and come back next time.

Listen to the podcast. Find it on iTunes/iPod and Podcast.com. Download the transcript. Get a copy of Glitch: The Hidden Impact of Faulty Software. Learn more about governance risks. Sponsor: WebLayers.

Transcript of a sponsored podcast on cloud computing and the necessity for automated and pervasive governance across a services lifecycle. Copyright Interarbor Solutions, LLC, 2005-2010. All rights reserved.

You may also be interested in:

No comments:

Post a Comment