We at Molnfront has since the start used many different cloud services (eg Cloudbees, Amazon EC2, NodeJitsu, Joyent, GAE, Jelastic, Heroku and Rackspace). The advantages are that they are robust, scalable and provides high availability at low cost. It is also less maintenance and ongoing operation.
Previously, if we wanted to have our own server to run an advanced dynamic internet services we rented a VPS of any vendor and installed a Linux server, we had full control but it was too much to think about in terms of configuration, updates to the OS, Security patches and daily operation.
Now we run all our services in the cloud and the difference is that we do not need to think about the operation and maintenance anymore and that we must be more careful with how we manage and optimize resources so we don´t get any runaway costs unnecessarily. The control of the OS and the hardware is slightly different but nothing that adversely affect the safety, servers in the cloud still consists of machines that are in a data center. On the plus side is the fact that the servers and OS are managed by professionals and machines are constantly upgraded and replaced and that the data, applications and instances are floating around between servers and data centers, and it is not possible to know exactly in which server someone's data is located, which makes hacking difficult.
What is Google App Engine?
Google App Engine (GAE also called App Engine, even the acronym GAE / J is used) is a platform as a service (PaaS) cloud service for developing and hosting web applications in Google's data centers.
Applications run in a sandbox and are spread over numerous servers. App Engine provides automatic scaling for web applications when load increases and allocate more resources to meet the needs that arise. GAE is free up to a certain level of consumed resources. A GAE account that has no payment activated will show an error message when the free resources have been consumed. With payment enabled you decide a daily budget for the resources consumed and the cost is from about $8 per month depending on the load.
GAE was released as a beta version in April 2008 and is from September 2011, in production version.
Reliability and Support
All billed High Replication Datastore App Engine apps have an accessibility SLA of 99.95% for a cost of $ 9 per month. App Engine is designed in such a way that it can sustain multiple datacenter outages without any downtime.
With a premium account support is included for $ 500 per account / month. Free support is available in the App Engine groups and Stack Overflow. We at Molnfront also offers paid support for $30 month exl VAT, which includes everything that is related to the actual web app such as backup, operation and errors in the content management system.
Dark clouds in the sky
When Google, Amazon, Azure or any other major cloud provider has a problem they get lots of attention as opposed to when a small obscure web hosting is affected by operational problems. It is not surprising, for example, Google has over 21 million websites (google: appspot.com). The last outage for Google was on 26 October, it had to do with the load balancers and affected 50% of the traffic to GAE.
Already in 1936, Alan Turing proved that it is mathematically impossible a distributed application is not going to encounter an error. If you have 1,000 servers each individually rated at 99.9% uptime, on average, one of those machines is always failing. So this is a problem not just Google, Amazon or Azure has to contend with. The only solution if you want to have as close as possible 100% availability, is to eliminate all the places that have only one source that can fail (single point of failure). The simplest solution is to use multiple cloud services and replicate data between them, and use load balancers that can quickly switch between the clouds. Some cloud providers (CSP) can offer it as an option, for example: Joyent, Cloudbees and Amazon. You can also create hybrid cloud solutions, consisting of both private and public cloud services.