If only Google App Engine had a SLA...

I am using Google App Engine for some time already, spent long time testing with small applications for my company, and even started a simple open-source library called Jobtracker, running on it, that helps tracking development costs through managing resource work time. With this one year of experience so far I can say that this is still an early stage of what is to come in the future of cloud computing, but I have high expectations (regardless the company).
Nevertheless, I still has my doubts on using such service for business purposes, as its reliability depends entirely on third-parties. Google has always tried, amazingly, to provide high-quality services with no cost to its customers, such as Gmail and even Google Apps, which are both very stable and mature. Even though the paid versions are guaranteed with a 99.9% SLA of uptime, I am most sure the actual number is much higher even for the free versions. Good for us, kudos for them.

Back to the real world...

The reality for Google App Engine, however, is not just there yet. I asked before directly to Google personnel about their expected SLA and current one, with no luck, so I decided to do a small investigation on their outages (openly reported here) and roughly estimate its value. You know, being a representative of such service and trying to push my own company to use it, I need data (I meant numbers) to prove it. So, here are my findings, considering one year up to now:

Event	Outage* (min)
March 02, 2009 06:45 PM - 08:45 PM	120
April 15, 2009 06:28 PM - 06:45 PM	17
April 20, 2009 09:30 AM - 03:45 PM	375
July 27, 2009 04:40 AM - 09:58 PM	318
August 20, 2009 01:30 AM - 02:00 AM	30
November 17, 2009 07:20 AM - 07:45 AM	25
December 09, 2009 07:45 AM - 08:00 AM	15
February 02, 2010 12:28 AM - 04:26 PM	238
February 24, 2010 07:53 AM - 10:30 AM	157
Total:	21.58hs

About 9 incidents in one year and a uptime SLA of 99.75%.

Note: Not all incidents represented completed service outage, instead they are mere rough values to measure the overall system stability, and may differ from an official source, that can analyze with care partial disruptions and present more valuable data for us customers.

Last thoughts

Of course we can not take this too hard on Google App Engine, considering it is a service free of charge, with lots of features, tools, documentation, etc. that can guide you or anyone (even without proper knowledge of python, django, MVC, XMPP, etc) to build simple applications in the cloud without any expenses rather than your own brain. This is already an achievement, and I am sure the service (which is growing rapidly) will only get better from this moment on.

The more reliable it gets, the higher are the chances companies and even people may find it as a trusted in-cloud solution to place their businesses or ideas, guaranteed that they can sleep at night without customer complaints or disappointments.

After all, the cloud and Google App Engine is what this is all about: ultra-redundant failure-proof solutions (at least at our end, as we don't have to worry about infrastructure, hardware, software, etc).