the Future of Work

Google and 21 Pages

by Danny Kolke
published on: 06/18/08 4:34:02 PM

As an early adopter of Web app technology, I routinely get pinged by people I know with alerts and stories. Yesterday I got pinged on the Google Apps Engine outage posting on Techcrunch and the headline states "Google App Engine Goes Down and Stays Down."

What do I think?

Contrary to the article I was quoted in a few weeks back in eWeek called "Mud-Slinging as a Service" I don't view myself as a mud-slinger and would like to speak to the pragmatic and practical assessment of what is going on.

When Google announced the Google Apps Engine, there were many commentators on what this meant for the market as a whole. Many have elaborated on the doom this spells for platform providers as Google really gets into the game. I have three words for these doomsday prognosticators:

Yahoo, Siebel and Microsoft

Granted at the drop of any announcement by Google or Microsoft, there are those that consider the rest of the free market doomed. I like to remind people that the history of the marketplace over the past 10 years tells us a different story. When I started Etelos in early 1999, Yahoo owned search, Siebel owned CRM and Microsoft was untouchable. In fact, I was told that "whatever you do, don't say you are a platform."

Well, of those three statements that were very true in 1999, none are true in 2008. Yahoo doesn't own search, Siebel is not the dominant player it was in CRM and Microsoft is being challenged on many fronts and it's debatable whether they are the platform of choice %u2014 especially on the Web.

I See Three Things: Validation, Confirmation and 21 Pages

The validation for me is that the business of supporting applications on the Web is hard stuff. It takes more to keep developers happy than running an application platform with limited services and letting developers write to a limited set of exposed APIs. Creating value for developers requires giving them more control, stability and value to make it worth their while. (I don't want to go into my Opportunity Computing pitch all over again here, so see this post.)

The confirmation for me is about the strategy by which these services need to be made available. I believe that the biggest problems with architecting a Platform as a Service offering is how to make your services available. Some providers are using virtualization and calling it good. Others are providing Web-based IDEs and no access to the code or database. Some are providing interfaces to write code to an API that talks to the database as a service... and you can continue on with the list.

Our approach was to enable a migration path for conventional development to embrace new distribution and services. We wanted to support the business rules and allow Web developers to focus more on app features... but without forcing them to learn a new way of coding.

As a developer, do you want to know where your app is? Interesting dilemma for the cloud technologist because some believe that you don't care, they say "just save your data with us and we will take care of it." Hmmmm.

To support portability of existing apps, accessibility to new services and greater control, we invented a different model for code distribution, application synchronization and value added services. I feel that this is confirmation that it's the right approach.

The reason I mention 21 Pages is because I was recently asked to comment on my thoughts on the notion of 99.999% uptime for Web services. If you have 21 pages that take 15 seconds or longer to load within the course of a year, your service is now below the threshold of 99.999% available. Most people don't look at availability of service when talking about up-time, but availability is hard stuff. The standard to the business user of an app is not just that servers are up, but if their application is usable.

Summing it up

As I move on to my list of things to do at Etelos, I don't have any mud to sling. I respect what the providers in the market are doing and their different approaches to advancing the Web as a platform. I also have a lot of empathy about what the business user goes through when they are not available to access the Web app they want to use. Naturally, I am biased in my opinion about how to do it. So take that into consideration in my comments here, but I think Opportunity Computing is the way to go.

Good luck ;)