Thinking Out Cloud

Today Rackspace announced it is open sourcing its cloud platform and releasing it as a project called OpenStack together with others. The code should be available for download (if not now then in a few hours) on http://OpenStack.org. From the official press release:

OpenStack will include several cloud infrastructure components, the first being OpenStack Object Storage, a fully distributed object store based on Rackspace Cloud Files, available today at OpenStack.org. The second component, OpenStack Compute, will be a massively scalable compute-provisioning engine based on the NASA Nebula compute project and Rackspace Cloud Servers technology, available later in 2010. Any organization will be able to turn physical hardware into massively scalable and extensible cloud environments using the same code currently in production serving tens of thousands of customers and large government projects.

In addition to NASA's contribution of Nebula, a number of other companies have lined up behind the project. AMD, Citrix, Dell, CloudKick, Intel, Limelight, Puppet Labs, RightScale and Zenoss all participated in an OpenStack design summit hosted by Rackspace last week in Austin.

Back in February I wrote a blog post titled Rackspace: The Avis of Cloud Computing? in which I analyzed how Rackspace is competing with Amazon in the Infrastructure-as-a-Service space. I ended the post with the following paragraph:

Finally, one more factor will play an important role in the success of each of the cloud providers, and that is the ecosystem that evolves around their products. That means management, automation and monitoring tools (e.g., RightScale, CloudKick, enStratus), Machine Image support (such as the AMIs from Red Hat/Jboss on EC2) and other ancillary products and services that form a "whole product". Amazon has the lead here as of now, but Rackspace seems to be working hard to catch up.

And indeed today's announcement appears to be the culmination of a lot of that hard work. A smart move by Rackspace that encourages the ecosystem even further.

It also looks like we are entering the "Stack Wars", with the two leading IaaS cloud stacks being Amazon's (with Eucalyptus as the private cloud version of the Amazon stack) and now OpenStack. It will be interesting to see how VMWare, Microsoft and Google all respond to this move.

This will clearly have an impact on the cloud platform startups who were going for an open source strategy, namely Enomaly and Cloud.com. Rumor has it that the latter are already planning to support OpenStack and base their product on it. [Update: In the comments below, Peder Ulander from Cloud.com confirms their participation in CloudStack].

Also of interest is how this will affect proposed standards, such as OCCI.

As cloud computing matures -- meaning it is being used by increasingly larger companies for mission critical applications -- companies are shopping around for cloud providers with requirements that are more sophisticated than merely price and ease-of-use. One of these criteria is performance.

Performance has consistently been one of the main concerns enterprise buyers have had about cloud computing, as indicated from the chart of the responses to a survey conducted by IDC in Q3 of 2009.

To address this concern, and help potential cloud users in selecting their cloud provider, a number of research, measurement and academic groups have initiated efforts to actually measure and compare the performance of various cloud providers under a variety of circumstances (application use case, geographical location and more).

Here are some of the more interesting performance benchmarks out there today:

Compuware Gomez: CloudSleuth

Still in beta, Gomez CloudSleuth is likely to be one of the more important reference points for customer and media cloud performance testing. Gomez has developed a benchmark Java ecommerece application and measures the end-to-end response time of various cloud providers and locations. The tests are run from 125 end-user U.S. locations in all 50 states and from 75 international locations in 30 countries and are conducted 200 times per hour. Gomez is planning on adding the ability to benchmark a user’s own application.

CloudHarmony

Although just a startup, if it succeeds CloudHarmony is likely to be an important resource for customers and the media for evaluating cloud provider performance. CloudHarmony has a service called Cloud SpeedTest, which allows users to benchmark the performance of a web application across multiple cloud providers and services (servers, storage, CDN, PaaS). This service is currently in beta and is quite simplistic, but CloudHarmony is working on a more sophisticated version with additional features.

In addition, the CloudHarmony staff conducts a variety of performance benchmarks for specific scenarios such as CPU performance, storage I/O, memory I/O or video encoding and publishes them on their blog.

UC Berkeley: Cloudstone Project

Cloudstone is an academic open source project from the UC Berkeley. It provides a framework for testing realistic performance. The project does not publish as of yet comparative results across clouds, but provides users with the framework that allows them to do so. The reference application is a social Web 2.0-style application.

Duke University and Microsoft Research: Cloud CMP Project

The objective of the Cloud CMP project is to enable “comparison shopping” across cloud providers -- both IaaS and PaaS -- and do so for a number of application use cases and workloads. To that end, the project combines straight performance benchmarks as well as a cost-performance analysis. The project has already measured computational, storage, Intra-cloud and WAN performance for three cloud providers (two IaaS and one PaaS) and intends to expand.

BitCurrent: The Performance of Clouds

BitCurrent conducted a comprehensive performance benchmarking study commissioned by Webmetrics and using their testing service entitled The Performance of Clouds. The study covered three IaaS providers (Amazon, Rackspace and Terremark) and two PaaS providers (Salesforce.com and Google App Engine). It measured four categories of performance: raw response time and caching, network throughput and congestion, computational performance (CPU-intensive tasks) and I/O performance.

The Bit Source: Rackspace Cloud Servers Vs. Amazon EC2

According to its website, The Bit Source is an “online publication and testing lab”, which appears to be a one-man show and may not play a significant role going forward. It conducted a one-time benchmark comparing Rackspace Cloud Servers and Amazon EC2 performance. 

What are your thoughts about cloud performance and these benchmarks? Please share in the comments below.

[P.S. I added a new category to the blog called "Shopping the Cloud", which will include posts that discuss various aspects of comparison shopping for cloud providers.]

In APIs and the Growing Influence of Developers I wrote about three general eras in software business models: High-Touch, Open and Cloud.

In the so-called "open era" we saw many startups disrupt an existing market by simply taking advantage of the open model -- and specifically open source -- with either the (more common) "open core" approach or the "open crust" approach.

So the idea was (and to some extent, it still is) that you pick any established software product category and you come out with a similar product that freely provides the source code and almost always makes available a version of the product for free download.

Now, we are in the midst of the same trend but with cloud computing. Just pick any IT product category and provide it as a service in the cloud. As expected, when the trend started there were many skeptics who consistently said that for THIS product category, cloud computing won't work (whether software-as-a-service, platform-as-a-service or infrastructure-as-a-service, as the case may be).

So 10 years ago when Salesforce.com just started with their SaaS CRM app, people said that enterprises would never put their most sensitive lead, opportunity and sales information "on some web site." I remember reading for the first time the news that Salesforce.com landed Merrill Lynch as a customer. Sure it was a tiny deal made by some small division within the company, but I knew it was all over.

A similar thing happened not too long ago with Zoho landing a deal with GE, and since then both Zoho and Google Apps have won many deals with large enterprises replacing Microsoft Office with their SaaS offerings.

In the case of productivity apps, such as Office, Google Apps and Zoho, the naysayers said that the SaaS offerings are severely lacking in features. They neglected, however, to realize a few things:

• With a SaaS product, rolling out features is an incremental low-effort process, so that new features will be rapidly added
• For the vast majority of users, and the majority of uses, a relatively rudimentary set of features are good enough
• The SaaS offerings had a new set of features that desktop-installed applications would find it extremely difficult to replicate, for example, collaboration and sharing, automatic backups in the cloud, accessibility from any computer, etc.

I am increasingly convinced that there isn't any category of software for which the cloud is not a good delivery method. Here's another example:

I recently spoke to Tom Greenhaw, founder of Cashier Live. Cashier Live provides a software- as-a-service offering for point-of-sale (PoS): in other words, cash registers. What I found interesting about Cashier Live is that it requires non-commodity hardware and the software needs to interact with the hardware: open the cash register, print receipts, accept credit card and bar code scans, etc. Something that you wouldn't think is an optimal fit for SaaS.

But Cashier Live is doing well with hundreds of customers. They have overcome the technical challenge of hardware integration (for now using proprietary browser capabilities of Internet Explorer, but they are working on solutions for other browsers). And they provide compelling benefits, particularly for small and medium retailers who do not have an in-house IT staff and find it difficult to make big upfront expenditures on software licenses.

Do you have an example of another SaaS offering that is seemingly a bad fit for its category but makes sense to you? Please share in the comments. 

This year I will be hosting the QCon San Francisco Cloud Track. If you haven't attended QCon before, it's a high quality conference that takes place twice a year in London and San Francisco. The attendants are usually senior software development and IT executives, software architects and senior developers from enterprises and startups. The conferences are organized by the fine folks at InfoQ.

The theme of the cloud track this year will be Real-Life Cloud Architectures, with the focus being on production apps by larger organizations in the cloud. There will be five sessions all of which will take place on Friday, November 5, 2010.

The first presentation will be given by yours truly on selecting the right cloud provider (more details to come). Another presentation will be given by Randy Bias (@randybias) of Cloudscaling on the topic of building a private cloud. I don't want to give it away but Randy has some controversial views on the subject (and a lot of experience), so it's sure to be an interesting session.

Informal Call for Speakers

Although I have a few candidates for the other sessions, I'm still looking for speakers and panelists. In particular, I am looking for two things: a company that has migrated an existing production application to a public cloud and a company that has implemented a hybrid cloud application (meaning it runs both on-premise and in the cloud). 

If you would like to suggest yourself or someone else, please send me an email (if you have my address), leave a comment below, or contact me through Twitter or LinkedIN.

Hope to see you there!

In APIs and the Growing Influence of Developers I mentioned the bottom-up sales and adoption model. That post, naturally, was focused on developers, but I said that the model can apply equally to any type of "rank & file" users.

I was flipping through Salesforce.com CEO and founder Marc Benioff's book, Behind the Cloud, and came across the following passage, which is a good (and classic) example of how the model works well with products targeted at business users:

[Sales executives] Jim and David had been with us for only a week or two when an incredible opportunity unfolded at SunGard, a NYSE Fortune 1000 company. Cris Conde, who had just been promoted to CEO, was looking for a way to integrate the data systems giant's eighty different business units and saw CRM as the "unifying glue." The possibility of our biggest customer to date, and the thousand-user deal David had guaranteed, was right in front of us. At the time, SunGard's various divisions were working on myriad systems, but Cris noted that salesforce.com was the only one that was spreading virally. "The sales people were buying it on their own credit cards and going around their managers to purchase an account," he said.

That "vote" from the users was significant to Cris, and meeting with him to discuss our service's capabilities not only yielded our biggest customer at the time but also helped build a blueprint for selling to all enterprise companies.

Salesforce.com, as you may well know, went on to become the world's first $1 billion pure-play Cloud/SaaS company.

On June 17 I gave a lightning presentation at the Open API Meetup, along with Evan Cooke of Twilio and Oren Teich of Heroku. It was a really nice event, well-organized by Sam Ramji and Shanley Kane of Sonoa/Apigee, and hosted at Twilio's beautiful new digs.

The title of my presentation was "APIs and the Growing Influence of Developers".  The gist of my presentation takes many of the elements I have written and spoken about in the past and puts them together in the context of the meetup's topic like this:

Just as we have seen with B2C, we are increasingly moving towards a "frictionless business model" with B2B products, including those targeting enterprises. This means that as a vendor you are trying to remove any barrier from the customer to acquire and succeed with your product.

To succeed with the frictionless business model, you have to adhere to a couple of principles:

Zero Touch

The idea here is to look at every phase of the customer interaction -- from demand generation, through customer acquisition and product delivery to customer success -- and eliminate touch points which create friction. A zero touch model has several benefits: it removes friction for the prospect/customer, making it easy to buy and use your product, and it also significantly reduces customer acquisition costs (CAC) and operating costs.

In the demand gen phase you want to rely heavily on inbound marketing such as SEO/SEM, social media etc. You're then bringing the prospect to your web site where they should be easily able to self-educate (with short video tutorials, easy documentation, etc.), sign up for a free trial and get up and running with something that provides value within minutes. You also remove barriers by providing flexible pricing (such as pay-as-you-go).

In the last phase, which I'll call "customer success", you want to find ways to provide low-touch/no-touch support, upgrading and up-selling/cross-selling.

There is a lot more to say about this topic and it's probably worth several dedicated blog posts, but you get the idea. I will add that although it sounds fairly obvious, it is not trivial to implement successfully.

Bottom-Up

This principle is less obvious. One of the implications of the frictionless model is that it relies on bottom-up adoption. So you are targeting not the CIO, business executives or central IT, as in the old days of enterprise software sales, but on the rank & file, whether business users, IT ops personnel or developers -- depending on the nature of the product.

With this point, I wanted to establish the fact that developers are becoming increasingly important when marketing your software (something I have written about several times before).

Product Delivery and the Frictionless Model

Given that the meetup was about APIs, I focused my presentation on product delivery. Over the years, software delivery has changed, increasingly removing friction from the process.

We can generally think of three main periods, I'll call them: High-Touch, Open and Cloud.

High-Touch Era

In this "classic enterprise" era, folks at big companies who had an interest in a certain software product couldn't touch that product without talking to a sales rep from the vendor. In addition, pricing was never revealed on the vendor web site and contracts were always one-offs that required prolonged negotiations.

You can easily see how this is a top-down model. It requires the involvement of legal, procurement, management authorizations, budgets, etc., and that's only to run a proof-of-concept.

But the delivery of the software product itself was an extremely high-touch approach. Software vendors "darkened the sky"1 with professional services folks in an implementation project that would take weeks, months or sometimes years.

Open Era

But slowly, things started to change, especially as the Internet had started to become a mainstream, commercial medium. Some startups, such as WebLogic2 (before it was acquired by BEA, before it in turn was acquired by Oracle), started offering B2B software for downloads on their web site. And as they did that, they were targeting developers, not executives. And they had to make it easy for people to help themselves with documentation freely available online (a novel concept then) and with easy access to support via email (another innovation).

I intentionally called it open era and not open source era, because source code was just one element of making B2B software products more easily accessible to developers. but certainly, the open source movement is an important part of this phase.

It is because of the removal of these adoption barriers that developers had access to software like never before. First, they downloaded and played around with it, the started developing with it -- next thing you know it's in a production application. I call this phenomenon "enterprise creep-up"3, and it explains much of the success of open source software such as MySQL and JBoss.

Cloud Era

(Not to be confused with Cloudera) 

Eventually, as both the business thinking evolved and web technology matured, we came to the realization that software doesn't need to be "delivered" at all. Instead, it's provided as a service. With the concepts of Infrastructure-as-a-Service, Platform-as-a-Service and Software-as-a-Service, the idea expanded to raw IT resources, infrastructure software (OS, VMs, middleware, databases, etc.) and applications.

With this phase, almost any friction involving product delivery is removed. The user/customer does not need to download, install, configure and maintain software. There are no issues of OS and hardware versions. There is no hassle in upgrading -- upgrades happen behind the scenes and you benefit from them every time you log in.

This explains the massive success of Amazon Web Services, and particularly EC2. Developers had unfettered access to IT resources, which they could pay for by the drink and expense to their credit cards, without the need for any approvals. They started by kicking the tires, the prototyping, then developing -- next thing you know, there is a production application on the public cloud.

Software Delivery Models in the Cloud Era

So we now have a range of software delivery options, many of which I listed in Software Delivery Models in the Era of Cloud Computing. The list goes something like this:

• Professional Services 
• CDs 
• Downloads 
• Physical Appliances
• Virtual Appliances 
• Images for Internal Cloud 
• Images for External Cloud 
• Add-on/Plugin/Extension
• Software-as-a-Service/Platform-as-a-Service via Client 
• Software-as-a-Service/Platform-as-a-Service via GUI 
• Software-as-a-Service/Platform-as-a-Service via API 

Developers and APIs

As we are moving to a world where software is delivered as a service, if you want developers to use or integrate your service with their applications, the preferred access method is an API (and as other speakers mentioned at the meetup, a RESTful API at that). An API is a programmatic (i.e., automated) way to access the service, and therefore, the most frictionless delivery method of all (at least when it comes to developers).

Tying It All Together

So with all this background out of the way, the point of my talk was quite straightforward: As developers are growing in influence in the purchasing cycle of software, and as we are trying to move to a frictionless business model, and as APIs are the most frictionless approach to deliver your service to developers -- providing an API to your cloud service is critical.

What's your take on this? As a developer? As a vendor? Let me know in the comments.

Footnotes:

Footnote 1 - At the meetup I mentioned that I borrowed the phrase "darken the sky with professional services" from former Sun Microsystems CEO Scot Mcneally, who used to say this about IBM. That phrase always reminds me of the screaming flying monkeys from The Wizard of Oz so I showed the picture below.

Footnote 2 - As I mentioned at the meetup, a few months ago I exchanged some emails about this with my friend Weblogic co-founder Bob Pasker. As a homage to the Weblogic guys I showed a screenshot of their web site circa 1996, which I picked up from Archive.org:

 As a footnote to the footnote, someone on Twitter got excited by the use of Papyrus font!

Footnote 3 - At the meetup I had a slide titled "The Tale of PTB". PTB, or Ponytail Boy, is JBoss founder Marc Fleury's nickname for former Sun Microsystems CEO Jonathan Schwartz (the only Fortune 500 CEO in history with a ponytail). I talked about this blog post from Schwartz that describes the "enterprise creep-up" phenomenon nicely:

I was with a big customer of ours last year, and reading through my account briefing before the meeting, I knew we were doing well. An analysis of their download activity showed they were heavy users of Solaris and OpenSolaris, and they had a large internal community of MySQL users, as well. In the meeting, their CIO said "we love where Solaris is headed." I then asked if we could help with MySQL, and he said... "I banned it."

Not exactly a buying signal.

I was stunned. I asked, "why?" He responded, "Oracle is our global standard, and with 20,000 developers, people need to follow the rules." I said we had a very good relationship with Oracle, and started talking about how fast Oracle runs against our new Open Storage products. Until he interrupted me, "...but my ban failed." What? "We hire lots of people out of college every year, and they all come in knowing MySQL. All my prototypes are written to MySQL, and now I have a big base of MySQL apps I don't want to port, and a bunch of MySQL programmers I don't want to retrain. So I'd like a commercial relationship."

In a nutshell, that's adoption in action. Change in IT isn't just a top down phenomenon - it's more often bottom up*.

After another very long break, James Urquhart and I finally recorded another Overcast podcast and it's now live and you can listen to it here.

Here's a mirror post of the Overcast blog:

Listen to the podcast:

Download Show #13 in MP3 format

Show Notes:

 So after another very long break, we're back with show #13. This time our guest is Krishnan Subramanian who writes for CloudAve and can be found on Twitter as @krishnan. 

In this show we discuss the topic of the LAMP Cloud, which Geva started off in a GigaOm post, Who Will Build the LAMP Cloud?, and James responded to with Does cloud computing need LAMP?. In an indirectly related post, Krish wrote about the Relevance of Open Source in a Cloud Based World, so we invited him to join us in a conversation about the LAMP cloud: Does it make sense? Who needs it? And what's the role of open source software in the world of cloud computing?

We also talk about the adoption of Platform-as-a-Service and other topics.

Some of the companies, products and technologies mentioned in this podcast include: Amazon, Zend, Google App Engine, PHPFog, Salesforce.com, Engine Yard, Heroku and Microsoft Azure.

Follow us on Twitter: @jamesurquhart, @gevaperry

I have said before that the cloud computing paradigm will not just enable us to do the things we do better, cheaper and fast, but will allow us to do things we couldn't have done at all without it.

Enter CloudFab.

I didn't know this until I spoke to Nick Pinkston the founder of CloudFab, but it turns out that there is a variety of what are known as 3D printers. These are essentially machines that can manufacture digital designs into physical objects -- even a single unit -- using various materials such as plastic, metal or resin.

 These 3D printers are highly specialized to particular kinds of materials and processes and are very expensive. They often sit idle -- making their use even more expensive. CloudFab created a marketplace of companies who own these 3D printers and matches them with people who need to create objects out of their STL files (a standard CAD file format for 3D designs).

So in essence, you load your design file to the cloud, request the process and material, and after getting and accepting a price quote, you can have the object manufactured and sent to you. Pretty cool stuff, especially for people who want to prototype products or have small projects.

CloudFab is now out of beta. If you want to try it out but don't have an STL file, you can download one from Thingiverse, a community web site with open source object designs (cool in its own right).

PS Do we have a new term? Manufacturing-as-a-Service? Facbrication-as-a-Service? 

On Saturday, GigaOm published a blog post I wrote entitled Who Will Build the LAMP Cloud? Please read the full post, but I basically ask in it the title question and speculate on potential candidates and players including Google, Amazon, Microsoft, Heroku and Zend. 

But apparently there may be a more fundamental question: does anyone want a LAMP cloud? 

My rationale for asking the original question was as follows: There are platform-as-a-service plays for popular application stacks such as Java (VMForce.com and Google App Engine), .Net (Microsoft Azure) and Ruby-on-Rails (Heroku and Engine Yard), so why not the LAMP stack? After all, it is one of the most popular technology stacks on the web today.

But in the very first comment, Kiril Sheynkman questions the need for a LAMP cloud altogether, with the claim that LAMP is losing its relevance and being rapidly replaced by other components. For example, he lists MongoDB, CounchDB and Cassandra as alternatives to MySQL -- the M in LAMP.

James Urquhart picked up on this comment and posted Does cloud computing need LAMP? in which he writes:

I have to say that Kirill's sentiments resonated with me. First of all, the LA of LAMP are two elements that should be completely hidden from PaaS users, so does a developer even care if they are used anymore? (Perhaps for callouts for operating system functions, but in all earnestness, why would a cloud provider allow that?) 

Second, as he notes, the MP of LAMP were about handling the vagaries of operating code and data on systems you had to manage yourself. If there are alternatives that hide some significant percentage of the management concerns, and make it easy to get data into and out of the data store, write code to access and manipulate that data, and control how the application meets its service level agreements, is the "open sourceness" of a programming stack even that important anymore?

I agree with James that in a PaaS offering, the developer shouldn't care about the underlying infrastructure, and therefore, the L(inux) and A(pache) in the stack are less relevant. This brings me to something that may perhaps seem like a nuance, but may be important: perhaps my original question should have been: "Who is building the PHP cloud?" and not LAMP.

In fact, if you keep scrolling down in the comments to my GigaOm post, you'll see an interesting one from a fellow named Brian McConnell. Brian writes that he currently uses Google App Engine with Python. He goes on into some detail about the benefits of GAE, which are essentially the selling points of PaaS over IaaS. But he concludes with the following:

I’ll probably stick with App Engine and Python for a while since things are working well and it ain’t broke, so it don’t need fixing, but if I am required to migrate in the future, or start another project, I’d like to be able to use PHP with the same level of simplicity.

This anecdotal testimony gives some merit to the notion of a PHP-focused platform-as-a-service.

James' last question: "Is the 'open sourceness' of a programming stack even that important anymore?" is a good one, but orthogonal to the discussion about a LAMP/PHP cloud, in my mind.

One of the complaints I've heard from users of public cloud users, and especially of those who use Amazon EC2, is that the instances available are not provided in a wide enough range of sizes and jump from one size to another in steps that are too large.

The problem with this is that the user ends up not utilizing all of the available resources of the machine they are paying for, which is somewhat ironic given that one of the biggest benefits touted for cloud computing is better utilization (as you are supposed to be paying only for what you use).

Librato (Twitter: librato) is a start-up attempting to attack this problem with their newly beta-released Silverline product. I spent some time with Fred van den Bosch (CEO) and John Wernke (VP Marketing) and was impressed.

Once you deploy Silverline (which takes a couple of minutes) on your Amazon instances (or any other instances in internal or external clouds), it "sponges up" unused cycles so that you can run apps that are non-time-sensitive such as analytics apps or various batch processes. And the beauty is that you are doing this on a resource you were paying for anyway.

Check out the video demo below. I'm curious to know what you think of this and what you think could be interesting use cases. They are now offering sign-up to a free beta. 

Please share your thoughts in the comments.

 

A couple of weeks ago the good folks at Novell invited me to participate in a podcast they are sponsoring called CloudChasers.

You can listen to it here or on iTunes.

Here is the description:

Battle of the Public Clouds: Who is Winning? – April 22, 2010

Many prognostications about the public cloud focus on three key vendors: Amazon, Google and Microsoft. This week on cloudchasers, we’ll check out the numbers, platforms and competing visions as we look at each vendor’s place in the market today and in the future. Also, with so many companies jumping on the cloud bandwagon, are there others who are more appropriate for this list? Who do YOU think will be the winner? Will there be just one winner?

Host: 

• Matthew T. Grant

Guests:

• Geva Perry, cloud computing blogger, Thinking Out Cloud 
• Ross Chevalier, CTO, Novell Americas

In Integration-as-a-Service Heats Up I talked about the cloud integration space and mentioned three companies: SnapLogic, Cast Iron Systems and Boomi. Well, I guess I was right about it heating up because on May 3 IBM announced its plan to acquire Cast Iron. 

It's important to note the difference between the three companies. Among the three, Boomi is the only one that has taken a pure cloud approach and provides its product solely as a web service. Cast Iron had started as an on-premise appliance play (virtual or physical) and recently added Cloud2, which they describe as a "multi-tenant integration-as-a-service cloud offering." SnapLogic is a downloadable product that needs to be installed on-premise or deployed by the customer in a public cloud environment such as Amazon Web Services.

[For more on application delivery models see Software Delivery Models in the Era of Cloud Computing]. 

Besides writing this post to say "I told you so" :-), I wanted to call attention to the debate about the appliance versus the pure cloud service approaches. In a blog post entitled Why Integration Appliances Won't Scale for Cloud Computing, Boomi CEO Bob Moul essentially slams Cast Iron's approach (although he is very gracious about the acquisition). The title of the post pretty much says it all, but you can read the details on their blog.

To me this is part of a bigger debate that is going on, and that is the fate of the notion of internal clouds (infrastructure-as-a-service and platform-as-a-service environments that run in the company's own data center).

Several of the large vendors, such as IBM -- as well as many startups such as Eucalyptus -- are pushing the notion of internal clouds. And they are gaining some traction. From the customer perspective the idea of internal clouds resonates because they want to provide their development teams with the flexibility of public clouds such as AWS, but without the huge headaches of governance, security, migration, lock-in and other problems associated today with external public clouds.

And then of course, there is the idea of a hybrid environment -- where some aspects of IT are run in-house and others in a public cloud or as software-as-a-service, which is where an offering such as Cast Iron's comes in.

But at the end of the day, Boomi's Moul has got it right. It is a model that won't scale. "Pure" cloud will win this war. Internal clouds will be a temporary stop-gap during the next 2-3 years, but will eventually settle as a small niche.