What's the difference between a television and a movie theater? Are they the same thing? What about a television with broadcast and a television with cable? What about with TiVo? What about watching TV on Hulu? Are they the same thing? I could describe all these experiences at a high enough level and make them sound like the same thing. But they are not.
From a consumer standpoint, you might not care if you're getting your TV over an IPTV connection or a cable connection as long as the video quality is good -- just like you don't care about DSL verses cable broadband as long as it's fast enough. But there are definitely business values to these varying technologies that matter a lot.
Recently in a conversation with two of the smartest guys I know in the online ad industry, there were some rather disparaging comments made about cloud computing -- essentially calling the whole thing marketing hype. After all, "isn't the internet defined as many connected servers facing user clients with no single, central server? That makes the cloud simply a whole mess more of them, no?"
Well, actually -- no. Think of the internet as infrastructure. You can do lots of different things with that infrastructure, including cloud services. And there are really significant implications for online advertising -- and for the evolution of marketing in general.
Let's say you owned an internet ad startup that was building a new ad serving system.
Approach A: The old-fashioned way -- without using cloud services
Write your software and prototype it in a small lab environment "on premise." Build out a data center with dozens of servers, multiple databases, and massive amounts of storage for the log files. Let's say the load of the impressions is expected to be around 30 billion a month.
Now let's say the peak load for delivering those ads requires 200 servers, but the average load is 50 servers, and the lowest load requires five servers. That means you would need to have 200 servers to handle the peak load -- but that peak load only represents a small amount of time. That's quite expensive, and a bit frustrating to only have a tiny bit of utilization of your server infrastructure -- and to have the financial obligations and tax implications of the investment. And on top of that, the servers are moving closer to obsolescence every day. Keeping the whole thing up to date on newest versions of hardware, server software, and database software is complex and requires a lot of people to manage the systems.
Approach B: Building out on cloud services
Now let's say you went down route B, in which you build your technology on a cloud services platform like Windows Azure or the Amazon Elastic Cloud 2.0 (EC2). You build your software so that the cloud platform can dynamically balance the number of servers dedicated to the service based on the load experienced at that moment in time. The overall cost of starting the company is dramatically lower, and the total cost of ownership of software and services is dramatically lower.
Now let's say you're an enterprise with a huge internal IT data center that you use for managing your ERP, CRM, and other enterprise capabilities. And now you want to integrate your online marketing data into the marketing business intelligence systems your analysts use to figure out how to spend their budgets. But in this world -- as opposed to the volumes of data you deal with offline, where you're dealing with a gigabyte or two of data a month -- you're dealing with terabytes or even petabytes of data a month. And you've never handled anything on that scale with your corporate IT resources. No problem -- put the data onto a cloud storage system, then build your analytics capabilities in the cloud. You can operate your business just like before, but without having to suddenly build out a huge new datacenter and develop new capabilities for dealing with massive amounts of data.
Once you have your online marketing data in an environment that enables you to merge it with data from other sources -- say, the U.S. Census or health statistics or mapping data or location data -- you can start doing analysis in much more valuable ways. The same goes for applications.
Cloud services may not sound super sexy -- it may sound like a techie kind of discussion for marketers. But you should know that cloud services are powering many of the new technologies in the online advertising space. I haven't talked to a single startup that was started in the last couple of years that isn't using cloud services for a big chunk of its infrastructure. And these platforms will become even more powerful and compelling over time.
Eric Picard is the advertising technology advisor to the Advertising Platform Engineering team at Microsoft.
On Twitter? Follow Picard at @ericpicard. Follow iMedia Connection at @iMediaTweet.
