Wednesday, August 06, 2008

Development Tools For Cloud Computing - Two Paths

For cloud computing to take off, there need to be tools available that enable a developer to build and deploy an application without having to download anything to their desktop. This requires an on-demand development tool that sits on top of the cloud and provides a development Platform as a Service (PaaS).

There are two paths that a vendor can take to create a development platform for cloud computing: cloud-first or tool-first.
  • Cloud-first approach to PaaS: first build a cloud platform, then build a development tool that runs on top of it. This is the approach pioneered by Force.com and followed by Coghead and Bungee Labs.

  • Tool-first approach to PaaS: first build a development platform that is host-able tool (e.g., studio runs in a browser), then "push" that platform into the cloud. This is the approach taken by WaveMaker.
For Force.com, it made a great deal of sense to take the cloud-first approach. SalesForce.com already had a robust cloud platform and expertise in building proprietary development tools to create their CRM application. There was also no requirement to make Force.com work on any other cloud, because SalesForce is aiming to be the only cloud you will ever need for all your enterprise apps.

For most software vendors, however, the cloud-first development process has distinct disadvantages. First of all, it puts you in the data center operations business, which requires a very different DNA than software development. Next, it makes development itself difficult, because the cloud adds a level of indirection and complexity to all development tasks. Finally, you will be forced to do cloud port eventually to get to a SaaS cloud people want to deploy on, like Amazon EC2 or Google App Engine (assuming they ever exit the Python ghetto).

A tool-first approach to PaaS development is much more straightforward. You start by creating a host-able development studio (pretty much rules out Eclipse plugins) and do your build and test on standard hardware. After you have build a solid product, you add multi-tenancy to the studio and customize deployment for your cloud of choice (or use a partner like Elastra to do the deployment and administration for you).

A final oddity of the cloud-first vendors is that they have all delivered proprietary development platforms. This provides a "roach motel" level of lock-in - your logic and data can checkin, but just try moving them to another RIA or Ajax platform! Again, SalesForce can throw its 500-pound gorilla weight around and make the Apex language successful. It is hard to imagine, however, that 5 years from now people who have learned the Coghead language will be in more demand than, say, Java developers.

2 comments:

Daniel Dean Gutierrez said...

By Dan D. Gutierrez
CEO of HostedDatabase.com,

When we launched the web's first Database-as-a-Service product offering in 1999, the decision for interface tools back to our web-hosted database was simple: straight HTML. Our service generated all the HTML necessary to hook back into our system. It still serves our customer base well, but then we cater to the lower end requirements.

Christopher Keene said...

I think the bar is a bit lower for a database in the cloud than for a visual development tool in the cloud. People are ok with database tools that are pretty bare bones - on the other hand, they are used to the power of development tools like Eclipse for app dev, so I think are going to be more critical of dev tools in the cloud