Thomas Bittman » Amazon

Four Myths About Cloud Computing

Tom Bittman — Sat, 09 May 2009 02:00:03 +0000

I’m sure I could come up with more than four, but let’s throw a few out there that I continue to hear from people trying to understand the phenomenon. Don’t misunderstand my intention here – I believe cloud computing will be huge, especially for commodity services, especially for small businesses and start-ups, especially for new and innovative applications that will leverage massive scale and low barrier to entry. But…

Myth: There will be a “big switch”
Fact: There will be a slow migration (including development of private cloud services), the migration will take decades, and even then quite a bit of IT will stay in-house; in fact, most of the interesting stuff will be hybrid models, long-term.

The electricity analogy really doesn’t fit well. Unlike electricity distribution (using AC instead of on-premises DC), IT is evolving at an extremely rapid rate. The number of enterprises generating their own electricity in 1887 was miniscule compared to the number of enterprises generating their own IT today. Even so, the electricity grid did not take place over virtually overnight – it took more than two decades before centralized utilities produced more than half of the electrical production in the U.S.

In many ways, the invention of AC was like the invention of the Internet. So why hasn’t computing across the Internet replaced enterprise computing yet? In IT, the “distribution” mechanism has been in existence for more than a decade. What’s really changed are technologies that enable economies of scale and sharing. We should not ignore the fact that enterprises can gain some economies of scale themselves, internally, for example, through virtualization – and enterprises aren’t ignoring it. The evolution toward cloud computing is a multi-variable equation, and does not have an inevitable conversion of everything to cloud services.

Myth: Cloud computing is just an evolution of “fill in the blank”
Fact: Cloud computing did not appear out of nowhere. Some say it is just the next version of outsourcing. Some say the next version of the web. Some say the natural evolution of virtualization. I say it’s all the above. The web created the standards and connectivity needed to make cloud computing possible. But, economies of scale do not occur unless you have technologies at the back-end that enable efficient technology sharing – multitenant applications, virtual machines, parallel programming mechanisms, automation, etc. Sprinkle in a growing demand for speed in the marketplace, and the industrialization of IT (including increased commoditization of hardware and open source), and cloud computing – has been evolving for years. And still has a ways to go.

Myth: Only megaproviders will win
Fact: There are diminishing returns to economies of scale, there are many fragmented markets that have good enough scale for smaller providers, and innovation makes provider agility a critical offsetter to size. We’re not going to have a handful of megaproviders, we’re going to have thousands of providers, and it will be very Darwinian.

Myth: Cloud computing is about IT commoditization
Fact: While services offered in the cloud may be commoditizing, the usage of those services may not – new, innovative businesses, proprietary analysis of data in the cloud, etc. – new applications matter. In fact, innovative use of cloud computing services will be be a huge reason why IT does matter, and innovative use of IT will remain a critical business differentiator.

And be careful – cloud computing services are not always cheaper. Providers gotta make a living. Amazon’s recent introduction of Reserved Instances was both to help Amazon plan and manage capacity, and to lower the price to compare better when workloads are not dramatically elastic. There’s a reason that some startups born on Amazon created their own data centers as they got bigger and their businesses became more predictable.

Karl Marx Would Be Proud

Tom Bittman — Tue, 31 Mar 2009 00:13:44 +0000

But cloud computing doesn’t need a revolution – capitalism will lead the evolution of cloud computing.

With only a few word changes, the opening of the Communist Manifesto could be easily inserted into the Open Cloud Manifesto, announced March 30, 2009:

“A spectre is haunting the cloud – the spectre of openness and standards. Amazon, Google, Microsoft and Salesforce.com have entered into a holy alliance to exorcise this spectre…”

This essentially describes the fear and the motivations behind the supporters of the Open Cloud Manifesto. Supported by a number of vendors who do not have a major cloud presence, including Cisco, EMC, IBM, Sun Microsystems and VMware. Not supported by vendors who already have a major cloud presence, notably Amazon, Google, Microsoft and Salesforce.com. Notice a trend?

The manifesto is simple and straightforward – so simple that the six “Principles of an Open Cloud” can be summarized as “Don’t use market position to lock-in customers, and drive cloud adoption through standards and collaboration directed toward customer needs, not provider needs.” In other words, stop building cloud services and stealing our customers until we get our act together to build cloud-based solutions for them.

Clearly the principles laid out in the manifesto are noble ones, and worthy of pursuit. But we’ve seen this play before. Vendor successfully creates a new market space, competitors cry foul and demand open standards and interoperability. “We want some of your customers!” The only difference with cloud computing is this is all taking place very, very quickly.

In the end – sorry Karl – capitalism will lead the evolution of cloud computing. It starts with innovators who pave the way. Let Amazon, Google, Microsoft and Salesforce.com build their proprietary cloud computing services. We need pioneers before we need standards.

As this market matures, customers will demand interoperability, portability, and federation with enterprise private cloud services. It is in the interests of newcomers to the cloud computing market to work together. The longer standards for cloud provider interoperability do not exist, the longer the entrenched cloud computing innovators will own the market.

So I have no problem with the Open Cloud Manifesto. It’s driven by capitalism – and in the end, the market will benefit. But don’t expect much soon, and don’t expect Amazon, Google, Microsoft or Salesforce.com to jump on board anytime soon. In fact, I’m sure they will be bashing the Manifesto and standards efforts for quite some time.

Will Cisco Unify Computing?

Tom Bittman — Wed, 21 Jan 2009 22:30:40 +0000

Cisco’s public blog recently announced an architectural approach they call “Unified Computing”. There’s been a lot of speculation about Cisco moving into the blade server business and so forth. I think Cisco just made clear that in their view the network may be at the center of IT, but “unifying” it with computing and storage is critical. Cisco is going to enter new markets. Maybe, better said, the lines between the technologies and the markets are getting fuzzier.

Gartner first talked about this in 2001 as Real-Time Infrastructure. IBM focused on “On Demand” which has been morphing over time but is still essentially a core part of their strategy. HP had Adaptive Enterprise (and Infrastructure) which shifted toward Adaptive Infrastructure and is still core to them today. Microsoft had Dynamic Systems Initiative, which became Dynamic IT – their vision behind the architecture for internal IT and for their cloud offering, the Azure Services Platform. VMware has Virtual Data Center Operating System and vCloud. And on and on.

I hear Harry Nilsson singing: “Everybody’s talking at me, I don’t hear a word they’re saying…”

Here are the words: unification, virtualization, cloud, adaptive, real-time, simplification.

We’ll see what Cisco delivers, but what is apparent is that the comfortable sandboxes in which different IT vendors sat are shattering. Those words demand that computing become a much more flexible, unified fabric. Unified to deliver on service levels. Services-oriented. Selling components is not the name of the game – making it all work together is.

What is also apparent is there are many vendor attempts to achieve this, and they all bring their current strengths and products to bear to unify a portion of the fabric. I believe Cisco’s announcement may be “one large step for a vendor, one small step for vendor-kind”. It is safe to say this will be big for Cisco – and big for unifying networking and computing – but it may not be a huge state of the art shift for the industry. It is good to see Cisco aggressively joining the club of vendors pushing the state of the art in infrastructure forward, however.

Finally, what is apparent is that we have to get over the religious argument about whether cloud computing can only be used to describe “what Amazon delivers” or mega-monstrous-external-service-providers, or unknown and unknowable computing capability in the ether. There is huge industry energy pushing in the direction that will make internal computing more real-time, on demand, adaptive, dynamic, unified. Cisco is undoubtedly a part of that push. What was custom will become packaged, and we will see a growth both in the numbers of cloud computing providers and in the number of organizations that feel they are building “private clouds” to be used only by their internal customers.

Where “the sun keeps shining through the pouring rain.”

Virtual Cloud Privacy is Gray

Tom Bittman — Thu, 08 Jan 2009 22:26:10 +0000

What’s private? What’s public? What’s in-between? We’ve been talking quite a bit lately at Gartner about variations of isolation in a cloud computing architecture. It’s not black and white – it’s gray (aren’t all clouds?).

Private cloud computing requires that the enterprise have dedicated data centers and everything inside.

Public cloud computing assumes that the enterprise is using a service where everything is potentially shared with other users. Everything. Including transactions, data, analysis of the data – everything that took place or was stored as a part of the service (think Google search).

There are plenty of variations between the two – here’s a few:

Perhaps you don’t want to share the transactions and analysis. You’re willing to use/share public data, you’re willing to share software – but how you use the software is your business.

Or perhaps the data is yours. You will share a multitenant application and everything below it (think Salesforce.com).

Or perhaps the you don’t want to share the application – you want your own dedicated copy, running on your own dedicated operating system, perhaps running in a virtual machine (typical Amazon EC2 deployment).

Or, perhaps you want better isolation – you want your software to be the only thing running on your hardware at any one point in time. You may need more, you may need less, but when you acquire it it is all yours until you are done (typical horizontal scaling for web servers in the cloud).

Or perhaps you want even more isolation – the hardware is reserved for you and you alone (think Exchange Online Dedicated).

And in all of this, management could be dedicated to just your stuff, or it could be shared. Management could even be running on dedicated hardware – or shared. And there are more variations.

A term starting to float out there in the industry is “virtual private clouds”. Reuven Cohen probably invented the term, but his use is different from some of the uses I’ve heard from vendors. Beware! It will be really important to understand what is truly “private” and what is truly “shared” when vendors start to use that term and many others to discuss what they’ve got. I can already hear the drumroll for more cloud computing terms being used in fifteen different ways – or is that thunder?

Data Centers, Clouds, Virtualization and Vegas

Tom Bittman — Thu, 04 Dec 2008 03:40:10 +0000

What a crazy and exhilarating week in Las Vegas!

I delivered the keynote at Gartner’s 27th annual Data Center Conference this week. Here are four major points I made. Later I’ll post some key take-aways from hours and hours of conversations – lots of great insights from conversations between our analysts and more than 1,800 attendees in a very busy week.

The future of infrastructure looks like cloud. Cloud is definitely being over-hyped in the marketplace, and isn’t popping into existence out of nothing. Gartner has been describing a vision for the future of infrastructure for more than seven years called “Real-Time Infrastructure”. Amazon’s EC2 and Microsoft’s Azure are examples of real-time infrastructures being delivered as external services. They are showing what can be done when standardization, virtualization and automation technologies are leveraged heavily, and internal IT organizations should learn from these examples. Cloud computing has a long way to go – but it is and will be evolving quickly. Evolve to behave more like a cloud provider – which will also prepare you to leverage cloud services when it makes sense.

Virtualization is a modernization catalyst. Virtualization is the hottest topic at the conference – but our key advice is to look beyond simple consolidation and cost savings. Virtualization can be the catalyst to drive many fundamental important changes in architectures, processes and cultures. Even if short-term attention needs to be given to cost-savings, make sure you build a foundation that can be leveraged in a few years. Virtualization “unlocks” cloud computing potential internally and externally.

Sourcing will become dynamic and granular. Traditional outsourcing and hosting is not dead, but it is dying. Economies of scale are replacing skills, and speed is replacing static contracts. IT needs to take the initiative in building dynamic sourcing teams – people who understand both IT and business, and can make sourcing decisions a project at a time, and dynamically based on changing workloads, priorities and costs. On the other hand, cloud computing portends a huge shift in small business IT. Cloud providers will offer a low barrier to entry and economies of scale that are too compelling to ignore. To service them, a new industry of service brokers will evolve.

The time to build a strategic plan is now. Infrastructure demands, an economy in turmoil, technologies for virtualization and automation, and the emergence of cloud computing are all happening too rapidly to simply stay in react-mode. Infrastructure strategic planning is not an option, it’s a requirement to navigate change and survive. Success stories are pointing the way. The gap between the strong and the weak is expanding – don’t be left behind.

More later.

Does Azure Imply a Cloudless Sky?

Tom Bittman — Tue, 28 Oct 2008 04:47:13 +0000

Is Azure a hopeful name? An azure sky implies cloudless, doesn’t it?

No question, Microsoft would like cloud computing to go away. Selling gazillions of copies of Windows and Office, not to mention mere zillions of Windows Server and Server products has been a nice business. This whole cloud idea is ruining their bright, clear, sunny day.

Cloud computing is just too easy for developers. Amazon, Force.com and Google Apps Engine are platforms that are just there, easy to get started. Not to say that Amazon, Salesforce or Google have all the kinks worked out – but they sure have lowered the barrier to entry for a developer looking to build a global-class application on the cheap. The Azure Services Platform is Microsoft’s answer to their loyal .NET developers – they will provide the same low barrier to entry and global-class experience. (By the way, the acronym for the Azure Services Platform is really, really unfortunate.)

That doesn’t mean that Azure has the management ecosystem, the elasticity, the quality of service controls that developers need – but give Azure time. This is a long-term play.

Software as a service and cloud computing is also appealing – within limits – to enterprises. Cloudsourcing commodity software, short-term software and capacity peaks sounds great. Extending desktop apps sounds great. Moving critical business data off-premises is less appealing. Finding that right balance – that’s what most enterprises want. Microsoft’s Software Plus Services strategy is intended to respond to that, providing the spectrum and the hybrid offerings that respond to more conservative enterprise demand – and also retaining as much of Microsoft’s traditional business as possible. This is a pragmatic strategy, and Azure will help Microsoft deliver on it.

But, Azure doesn’t support existing applications – applications need to be targeted originally for Azure. With Azure, sourcing is not a runtime operations decision, it is an application design decision. The software and/or services decision is kinda hardwired at design time, which is unfortunate, and means that enterprises will need to look elsewhere for solutions to cloudsource some of their computing requirements.

But what these enterprises really want is help being more efficient internally. What are the Secrets of Azure (doesn’t that sound alluring? Gotta use that again…) that could help enterprises be more efficient and cloud-like?

What about this new Windows Azure “Cloud Operating System”? Sounds pretty cool to me. The only problem – this isn’t a cloud operating system at all. Windows Azure is just the operating system used on all the servers that are supporting the Azure Services Platform, and it is unclear that there is really anything significantly different at all in this operating system than packaged Windows Server 2008 running Hyper-V virtual machines. The real “cloud OS” is the unnamed “fabric” that does all the neat stuff, like monitoring applications, spinning up new VMs when needed, and detecting failed applications. Whatever that is, please box it and sell me some!

Without a doubt, what Microsoft needs to do is to maintain synergy between Azure and “traditional” Windows computing (Software Plus Services), as well as ensure learnings from Azure make it into the “traditional” Windows computing experience. If Microsoft wants the sky to stay azure, they should give enterprises as much of the cloud experience that they can internally – help them build “private clouds” – but always with the option to slide slowly and easily into the direction of Azure – at least easier than into the arms of Google.

Partly or Mostly Cloudy?

Tom Bittman — Thu, 09 Oct 2008 05:18:55 +0000

My head has been in the clouds for the past week or so – I’ve literally been working full-time on cloud computing. Cloud services are popping up right and left, and more are coming. Some of these are not cloud by our definition. But what about the ones that fit the definition, but barely? From a cloud customer perspective, what makes one service more “cloudy” than another?

I think there are two obvious dimensions to “cloudiness”: Service and Elasticity. There is a third less obvious dimension that I’ll talk about later.

By definition (ours, at least), cloud services are services-oriented, but how much? At one end of the spectrum, a cloud service is highly standardized, uniform, one-size fits all. A service, but not very rich – probably designed more for high-volume and reach. I’m thinking John Belushi on SNL saying “No Coke – Pepsi!” At the other end, a cloud service is highly customized to meet cloud consumer needs, perhaps with a wide range of quality of service options – like Burger King’s “Have it your wayyyy!” (OK, that’s just marketing, but you know what I mean). In the early days of cloud computing, most cloud services will tend to be relatively simple, with limited options. As cloud services mature, more and more personalization and choices will emerge.

Elasticity is also a requirement, but how elastic? At one extreme, elasticity might be slow, or come in chunks. It might be difficult, and require some intervention. It might reach some low-end or high-end scale limits. There might be a high barrier to entry, or high barrier to exit. Kinda sorta elastic. At the other extreme, elasticity is rapid, very smooth and granular. Entry and exit are easy. Sourcing could dynamically move from one provider to another. The nirvana of utility computing. Early cloud offerings have elastic qualities, but they also have a long way to go. With Amazon EC2, for example, you need to define the size of server you need. You can add another one, but it takes a little intervention. This too shall pass.

I think there is another dimension, but I am struggling to name it (please let me know if you have ideas). Let’s call it market, for lack of a better term. At one end of the spectrum, a cloud service is highly monolithic, built on a customized architecture, an integrated top-to-bottom solution, perhaps less innovative and dynamic. A relatively closed market. I discussed this kind of cloud service in my last blog post. The other end? A cloud service built on top of cloud services, an ecosystem of services working together, very organic and dynamic. An open market of federated providers, constantly changing. These are quite rare today, but I expect more to develop over time.

The early cloud service providers are building castles in the clouds. Frankly, in these early days, we need to start this market with a few successful castles. But I expect dynamic and vital markets to bloom around and away from the castles as cloud computing matures. I also expect those monolithic providers to evolve to interact with the market at many different levels – but not right away. Will salesforce.com always have their own data centers? Do they want to? What about Microsoft? Maybe now, but later, after the market evolves? We will see.

VMware Strategy Reaches for the Clouds

Tom Bittman — Tue, 16 Sep 2008 19:21:44 +0000

At VMworld today, VMware announced two strategic shifts that are as important as they are unsurprising. One pits VMware firmly against powerhouses IBM, HP and Microsoft in the competition for infrastructure control. The other extends VMware’s virtualization story into cloud computing. Are they overreaching?

The first, “Virtual Data Center Operating System”, is what Gartner defined as a meta operating system in 2004. A meta operating system is a virtualization layer between applications and distributed computing resources; it utilizes distributed computing resources to perform scheduling, loading, initiating, supervising applications and error handling. Essentially, an operating system that manages distributed resources. If you have hundreds or thousands of individual operating systems, take certain common functions and consolidate them for the entire pool of resources. Microsoft had a research project years ago that focused on just this – they called it BIG, for “billions of interconnected gizmos”. That led to the Dynamic IT strategy, but we haven’t seen Microsoft put much effort into redefining the operating system concept because of it – yet.

VMware includes in their concept what Gartner calls a service governor, which adds policy-based management on top of a meta OS. Combined, these two create what Gartner calls a real-time infrastructure. The service governor is the real challenge for VMware, which is one reason they haven’t called it out.

What is interesting is that VMware is finally describing a larger strategy that is completely competitive with IBM (remember the On Demand Operating Environment?), HP (Adaptive Infrastructure) and Microsoft (Dynamic IT). The strategy is credible, but there are many, many gaps that need to be filled. In particular, while VMware is strong in virtualization, they are very weak in service management. Regardless, it will be difficult for IBM and HP to miss the competitive threat (which, of course, they should have seen starting in 2001). This is the only natural evolution for VMware, but the road is littered with challenges.

The second strategic shift was the introduction of vCloud – essentially, services that make movement of virtual machine-based workloads in and out of the cloud easy. It takes programmers to leverage Amazon’s EC2. vCloud’s intention is to give infrastructure administrators the ability to make sourcing decisions not only within a data center, but to and from cloud providers who leverage VMware’s technologies. VMware also hopes to build a market of cloud computing providers that use VMware’s software to power their cloud computing architectures. While there will be a growing demand for software that enables cloud providers, VMware can only play if their pricing for service providers is much lower than VMware’s price today for enterprises.

There’s a lot more to discuss and analyze, but I’ll leave that for a future post.