I am writing from San Francisco today, where I am attending both the Web 2.0 Expo and Location Intelligence conferences. I have found that the serendipity of discovering real potential value in the concept of “Web 2.0” while developing our solution for B2B mapping a bit humorous. My original take on the Web 2.0 business was that it was all about social networking and advertising. However, our industry (the global mapping industry) is ripe for a true SOA solution, and we are trying to build something that will release the potential of both the internet and mapping beyond just the ability to share mashups. In order to accomplish our goals we needed to overcome some critical infrastructure hurtles in the development of a platform that allowed real internet commerce to proceed within the mapping industry. As I am preparing for both of these conferences this morning I thought I would begin to share some thoughts on how we are planning to build a SOA, which may be considered a Web 2.0 application.

The global mapping industry is a $4 to 7 billion a year market (depending on which report you read). It is a B2B industry, dominated by large investments in infrastructure (think satellites, airplanes, computers, software, and content), as well as large investments in highly skilled technicians. The data volumes are enormous; our own mapping efforts (at FERI) run upwards to 10 terabytes of mapping products per day, requiring multiple distributed processors just to generate the maps, which we then have to serve to clients and users in near-real time. WeoGeo (www.weogeo.com) is our B2B portal and server solution to rapidly delivery mapping products to end user customers.

Imagine building the computational and internet infrastructure to deliver gigabyte to terabyte size maps. A terabyte map takes 90 days to be transported over a 1.5 megabit per second link. WeoGeo has developed the technology to dramatically reduce this effort, but to service a global market of such maps would require mind boggling infrastructure support. Enter Amazon Web Services (AWS).

The initial beauty of AWS is in the cost structure, where we are paying for our computing time (EC2) and data storage (S3) on a pay-as-you-go basis. Our initial budgeted start-up infrastructure costs were ~$300,000 plus first year expansion ~$200,000. When we budget the same effort on AWS, its pro forma was somewhere between $10,000 to $20,000. AWS allowed us to spend our limited start-up dollars on developing the technology of WeoGeo, rather than buying and maintaining computers. But the initial beauty is quickly overtaken by something a bit more sublime. With EC2 and S3 our processing and storage requirements are totally scalable. The term scalable is so prevalent in today’s business press that it often loses its significance. However, scalable to us has very significant time lag and costs implications. Besides pitching to potential customers who have map archive inventories approaching petabytes, we are talking about a web services business that currently counts 200 million Google Earth users. If we are as successful as we hope to be, an exponential growth in business would rapidly overcome our abilities to assemble hardware, much less install and maintain servers to service the business. Our business requires scalability, with a capital S.

So we made a bet at the beginning of WeoGeo that a business model built on commodity computing cycles or elastic computing, as opposed to commodity computers would best enable us to handle growth in this industry. The fact that the upfront cost was cheaper was a bonus. The bet required focus, so we decided to make an all inclusive AWS service platform that required no outside data center processing or storage. For this to work, our web and data base services had to be robust and durable in a virtual machine environment that in it of itself might not be durable. It had to handle spiking (think “Digg Insurance”) and cyclic patterns in processing to assure up-time and optimize costs. It also had to address load balancing and stable IP addressing in an environment where the virtual machines’ IP addresses and domain name records may be lost.

With a lot of brain sweat and great interaction with the AWS team, we created an internal EC2 management solution that accomplished these goals. After some prodding by the AWS team, we have begun to offer one of these solutions as a product. WeoCEO (www.weoceo.com) is a management solution for stable IP address, fail-safe monitoring, load balancing, and auto-scaling of EC2 resources. Besides the insurance aspect of this solution that provides for robust ecommerce activities, the auto-scaling feature actually provides a tremendous cost savings over daily and seasonal cyclic usage patterns. We look to providing an extension of the WeoCEO services for durable database operations in the near future.

In short, the mapping industry is competitive B2B market that has high infrastructure costs to support large processing and storage requirements. WeoGeo has created an SOA on AWS that will allow for the unleashing of huge volumes of archived mapping products to create a geospatial information exchange that will scale from the smallest to largest users. While cost containment will certainly be a key component to our viability, we believe that quick, reliable, and scalable service will be more important to our eventual success.

Adena Schutzberg at Directions Magazine (which is hosting the Location Intelligence conference) has indexed a podcast to be available on April 17^th titled, “Is Web 2.0 Mapping “Dead”?”. All I can say is that we don’t think so (and I sure hope not).