We're now at a point where all things "cloud" are sure to be the darlings of the dance in technology circles.  The nice folks over at ReadWriteWeb reference the Gartner 2010 Hype Cycle report  and concisely sum up what many of us already knew.  Cloud based computing is dancing right up there at the peak of inflated expectations. 

Based on what you read online, cloud computing will solve all of your IT woes, cure cancer, stop global warming, and iron your favorite dress shirt (light starch please!).  But how long until the cloud rolls over that peak and starts its downward slide?  Let's give it a little nudge shall we?

Over the past several weeks, and particularly subsequent to doing our ESRI/AWS webinar for Directions Magazine, we've been privy to lots of questions and conversations regarding mass movement of existing infrastructure to the cloud.  "We gotta get us somma that cloud!" Don't get me wrong, there is a bona fide list (albeit a short one) of business cases for organizations to leverage an entrely off premise solution for their organization.  For most mere mortals however, this is clearly not the direction they should be heading.  Cloud-based solutions are certianly not a silver bullet.  Just like any other large scale change in infrastructure, mass migration to a new environment will be highly disruptive, costly, and high-risk, particularly for LOB users.  All of these challenges are born for marginal overall gains in productivity and cost savings.  Infrastructure depends on actual humans for its care and feeding no matter where it sits.  In an off-premise scenario, all you've cut the cord on, so to speak, is the hardware.  You still have to manage connectivity, provisioning, storage, backup and recovery, and security.  Oh and by the way, you still have to manage all of your on premise systems as well.  Sounds fun!  So in the spirit of not selling snake oil, we actually propose an adoption strategy that is more measured and pragmatic, that manages risk and builds upon success.

Step 0: Can/Should we even try?

In many organizations there will be institutional or policy barriers to leveraging cloud based computing.  The idea of not being able to touch and feel that data server in the next room still gives many IT professionals fits.  In the GIS industry in particular, the idea that "spatial is special" tends to fade slowly. In addition, the prospect of a variable bill every month is not often looked upon kindly by the accounts payable department.  Federal, state, and local government organizations for example like to know exactly what they're paying when. 

In addition to the question of "can", there is the all important question of "should".  If IT provisioning in your organization is relatively painless, if performance is acceptable, and if maintenance issues are few and well understood, it probably makes little sense go mucking about with a good thing.

If, on the other hand,

• you require periodic scalability for LOB or mission critical apps and data beyond the capacity of your IT infrastructure
• if you've got competent bodies in place to manage the off premise resources
• if it takes you a year to provision a server and six months to install an app that took three months to build
• if organizational structure and policies permit

the cloud is certainly an option for you.

Step 1: Lay a foundation for small victories

The typical organization should begin by setting up an easy way for early adopters to leverage cloud based computing.  From the GIS perspective you'll want to secure licensing from ESRI, setup terms with Amazon, and create a streamlined way for various groups in your organziation to provision the needed AMIs.  The goal and main benefit in this scenario is to facilitate rapid development and deployment of focused applications by removing barriers presented by IT infrastructure challenges.  Allow groups in your organization to innovate in this new environment and you'll be surprised what they come up with.  Once you've had a few successes with this approach, proceed to step 2 if your analysis reveals a business benefit to you.

Step 2: It's all about services

Now that you've got some folks in house that are comfortable working in the off-premise infrastructure, consider setting up cloud hosted geoservices that are typically leveraged by multiple groups or large numbers of users.  A state government might consider working with ESRI and providing state wide base data layers for the ESRI Community Basemaps initiative as well as publishing map and data services that provide statewide base data layers.  Services in EC2 with a bunch of tile caches in S3 is a no brainer.  Begin by providing periodic snap shots of relatively static data with a gradual, measured migration to automated data syncronization between on-premise systems and your cloud infrastructure.  Once you're comfortable working in this model, do another gut check and if it makes sense, move right along to step 3.

Step 3: Wade in with both feet

So you've got your small focused apps and a healthy suite of GIS data services running happily in the cloud.  Now it's time for the next step...publishing organization-wide and public facing applications to the cloud.  A private sector company might elect to run an LOB app in Amazon, while a state government might decide to push statewide web mapping applications.  These migrations should be done only as they make sense.  Migrate an app as existing hardware comes up for replacement, or as new systems come online.  Always keep in mind that applications that require operational data from real-time or near real-time systems will likely need to make calls back to the on-premise infrastructure.

A fool and his money are soon parted

The basic message we're trying to push as it relates to cloud computing in the GIS industry is that the cloud, no matter how unique and nebulous and wonderful it seems, is not exempt from the technology hype cycle.  A more measured approach that balances risk with small initial victories and careful cost/benefit and risk/reward analysis for the organization will provide greater long term benefit.  Given that you're committing to the cost of an ESRI ArcGIS Server license as well as the monthly costs of living in the cloud, those dollars should be used intelligently where they will do the most good. The whole point of cloud based infrastructure is to make life easier, not to create the next GOS, NSDI, super-delux infrastructure mess...it's the same mess, you've just put it in someone else's living room.  Baby steps...small victories...and build gradually.