Web based systems typically require an investment in hardware, hosting and management to get them off the ground once they’re beyond a prototype. That is starting to change with things like Google’s App Engine being released.

The Google App Engine allows developers to create applications using Google’s framework, test them locally and then upload the applications into the Google virtual computing cloud (the same cloud infrastructure powering Google’s own applications). It provides integration capabilities with Google’s applications, build in online storage and a variety of other features to speed up launching an application.

While this sounds like the silver bullet of web apps, there are a number of reasons some companies won’t be using it.

First off, building a SaaS (software as a service) system that you’d like to make money with may not be possible (Google user account management is included but the capability to charge users and restrict access is not yet). Second applications are limited to specific technologies (they must be written in Python). This will make repurposing existing applications much more difficult and Python may not be the language of choice for most developers.

Google’s App Engine is a great first step into making systems scalability a simple infrastructure issue that developers can ignore but it will need to evolve to be deeply useful to most organizations.

Most systems (especially web applications) are built with very little thought about the post implementation life cycle of the system. One of the key components to staying ahead of the growth curve of a system (and quickly diagnosing problems) is a monitoring system that includes key system metrics.

Ping monitoring a server to determine if it’s alive is helpful to identify critical failures but does very little to tell you in advance when you’re going to run out of disk space or need to upgrade the server. The monitoring system should include key server level parameters (hard drive free space, memory used, CPU usage, network interface bandwidth, disk activity) as well as application specific items (numbers of users, page views, transaction counts, queue lengths, etc.) Having this information available for trending analysis makes it very managable to plan for system upgrades well in advance of the system going BOOM.

Identifying those key system metrics up front and building them into the application initially will be much more cost effective than try to retrofit them into an existing infrastructure. Putting in place a simple web based dashboard that includes graphs over time of those metrics allows any member of the technical or operational team to identify potential pitfalls.