I attended the Do it with Drupal conference in New Orleans last week.  I enjoyed the content, the people and the city very much.  Somehow, it was my first trip to New Orleans and I was overwhelmed with restaurant recommendations for the two nights that I was actually there.  Especially since I only managed to only eat part of one dinner anyway.  But, I can highly recommend the gumbo at Nola!

Acquia held a sponsor’s presentation during the lunch on the first day.  Bryan House presented on the company and their products/services.  In addition to providing a supported version of Drupal, Acquia also has a useful search module.  Many Drupal users have installed Apache Solr for search but that doesn’t give any context to the relationships between the nodes or allow for filtering.  Bryan presented the three main lines of business:  Acquia Network (support/search), Acquia Hosting and the new Acquia Gardens (Drupal as a sevice).  They also have a free version of their Drupal distribution that leverages the Drupal core and the modules that are most commonly used.

The Drupal community is very vibrant and growing.  It’s working to better understand how to scale a web application.  The installation itself is fairly straightforward:  a database + webserver with php.  However, complexity comes with scale and additional features.  There were a lot of references during the conference about the need for separating tiers in the architecture and inserting caching layers to help scale.  The MySQL tier is very important and needs to be architected properly.  It doesn’t seem like many users have looked at memcached but that will happen quickly as larger sites get built.  There is also a need to size and scale the app and web tier.  There were mentions about substituting apache with nginx or varnish (reverse proxy can be a huge help).  There are probably other open source tools that can help, but haven’t seemed to gain traction yet.

Congrats to the Lullabot team for a terrific conference!

I was reminded of this when I read an interesting blog post from Thomas Bittman last week.  Thomas is an analyst with Gartner who wrote about his recent experience at the Gartner Symposium in Florida.  He says that 75% of his audience raised their hands when asked if private cloud computing  was going to be a core strategy of theirs.  That’s a much bigger number than he expected and definitely larger than I would have expected.  Of course, that was a session on virtualization so I have to wonder how many attendees freely substitute by using “cloud” for “virtualization”.

There is certainly much confusion over how to describe a cloud service offering.  Is it “Infrastructure as a Service” or “Storage as a Service”?  Perhaps “Software as a Service” or something else entirely?  It seems that the only thing in common is that is’s some sort of service.  There are many different cloud offerings and it’s often defined by which cloud service is being used.  Which isn’t always clear, even Amazon Web Services is sometimes referred to by specific services (like EC2 or S3) and sometimes by the generic AWS.

It may be easier to describe virtualization first, although for many people the default definition is VMware.   The simplest definition is a logical separation of compute resources, which could describe anything from LPARs to LDOMs to Zones.  I first started using virtualization as a technique for server consolidation when server sprawl started to take over the datacenter.  It was not only a way to increase utilization but saved rack space and money.  Most modern hypervisors are more efficient and virtualization is a key tool in planning any infrastructure.  It’s now possible to either collapse layers of an architecture into a single hardware system or to combine less utilized applications into a more robust system.  Advanced feature of virtualization software do tend to make them more cloud like, but it is different at a basic level.  Clouds will use some form of virtualization but virtualization by itself does not make a cloud.

Clouds are different and somewhat nebulous (sorry couldn’t resist).  The key characteristics for a cloud infrastructure are an on-demand, virtualized infrastructure that used and billed as an on-demand resource.  This has many implications both for a user and a service provider.  From a user perspective, it’s important to plan for resources that will go away but can be used very flexibly.  It’s a very different approach than a long planning cycle followed by a datacenter installation.

The major differencce between virtualization and clouds is that a cloud becomes a service platform and the user doesn’t know or need to know about the platform.  It’s a resource to be utilized.  Virtualization is a component of that service but not the chief component.  The cloud doesn’t become complete until it becomes an on-demand, reliable, scalable service.  That sounds like a very appealing cloud to me.

The customer presentations were very interesting, ranging from Sony Music’s commerce sites to a ticker data analysis application from the NADSAQ.  I liked the new features from the Amazon VPC, especially the ability to extend a corporate networks IP addresses into the cloud to access compute resources.  I think that it’s a first step towards a public/private hybrid cloud.

Throughout the presentations, the customers extolled some of the customary cloud benefits like quicker time to market and pay as you go.  Werner Vogels talked about four key characteristics for AWS:  scalable, cost effective, reliable and secure.  Scalability in the cloud is not just how big, but how quickly can additional instances be provisioned.  Indeed, provisioning time is a key requirement for their internal processes.  Of course, there was a key difference in metrics between Amazon and a traditional enterprise.  Amazon had already reduced provisioning time to 6 hours, which still wasn’t quick enough for developers to release resources after they were done using them.  They decided that provisioning time had to be measured in minutes to make if efficient enough for a developer to freely put resources back into a pool for reuse.

It’s easy to contrast Amazon’s benchmark to most of corporate America where provisioning time is measured in months, possibly weeks but usually not even days much less hours.  This metric alone is certainly is a large contributor to the success of AWS in the enterprise.  Amazon is getting more and more examples like the NYTimes image archive where the cost alone for using their internal IT department would have been prohibitive.  They not only saved money but completed the project in days not months.  These are business owners who are taking ownership of the technology project in order to make a project achievable.

I think there is another tangible benefit to enterprises leveraging cloud resources for a new project.  There is no legacy software in the cloud.  Cloud computing infrastructure is the new greenfield for architecture choices and technology selection.  There are many success stories about an enterprises ability to be nimble by choosing external cloud resources and completing a project cheaper and quicker.  But one of the panelists last night said the key was to leverage this opportunity to start fresh and design an architecture to be nimble.  It’s a very different approach then designing only for reliability or high availability.  That doesn’t mean that an architecture should be resilient and it needs to meet the applications business requirements.

This can be a tremendous benefit to using cloud resources for a new project.  Cheap and fast have just become the barrier to justifying a project.  The real justification will be to select technology and design an architecture that will be nimble, allowing you react quickly in the future.  That’s a benefit that allows for growth not just savings.

Given the current economy, I don’t think that was a surprise for anyone in the room.  However, they did have some good suggestions on managing the RFP process and suggested that it is a best practice for anyone making a significant purchase.   I was surprised at some of the concessions that they’d acheived for both modified terms and conditions, as well as, services and licensing costs.

I was also interested to hear that they had clients asking for significant services commitments to help derive requirements as part of the RFP process.  Business requirements are not always a part of the RFP process and can lead to incomplete responses if it’s not clear how the technology will be used. The user perspective as part of that process was good to hear.

Another surprising comment was how little consideration was being given to open source software and cloud computing.  Some companies are using open source as a way to pressure commercial software vendors but few of the law firm’s clients were actively pursuing an open source strategy.  There may be some negative correlation to enterprises who use law firms versus those enterprises that are actively deploying open source.  But there seemed to be little enthusiasm for open source in the room.

I’ve seen many large (and small) enterprises deploying open source software strategically.  Depending on the business requirements, companies may only deploy open source for certain applications but overall there’s a much larger penetration than this group was willing to admit.  I’ve seen a  new “best of purpose” approach, where is heterogeneous software adoption is encouraged.  It’s better to manage multiple applications to get a more efficient infrastructure than trying to force fit a single application to do things it’s not well suited to do.  For example, an enterprise may have Oracle, MySQL and Hadoop for different applications and different types of data.  Admittedly more difficult to manage, but more efficient overall.

I’m sure it correlates with the open source adoption but there was also little testing or adoption for any type of cloud computing strategy.  While this will have an impact on the software licensing, the key is finding applications that are appropriate to deploy in the cloud.  Many enterprises are looking at cloud resources as an alternative to building or scaling infrastructure.  As this matures, it will become a much more prevalent choice.

It was a very interesting session and I appreciate the opportunity to hear different perspectives and how they view the process.