Comparing Cloud Platforms

Apr 1, 2010 Posted on Apr 1, 2010 in Cloud Platforms, Featured, Real World Azure, Strategy | 0 comments

Comparing different cloud platforms is almost like comparing cars from different categories. A sedan, a minvan or a 2 seat roadster will be the right choice depending on what dimensions you want to optimize for.
The offerings from Amazon, Google and Microsoft – the major players in this domain are distinctly different today. And, the size, ease of handling and maneuverability differ with each.
For instance, Amazon offers you complete freedom to install and maintain the operating system of your choice. But the Google and Microsoft platforms do not provide this luxury. But then who wants the luxury of maintaining an operating system environment? Most of us are interested in the end-result: running our applications.
So, how do you choose a platform that suits you best? Perhaps the best metric would be to assess how efficient the cloud platform is in replicating your in-house infrastructure environment, sans the burden of managing the hardware. Here again the task is not as simple as it looks. Here’s why:

Amazon provides the best options in terms of flexibility. Salesforce.com on the other hand offers only applications which have undoubtedly stormed the cloud with its popularity. (You must notice that applications provided by salesforce.com are niche applications meant for a small audience). Custom applications can be built on the force.com platform but they are native to that platform with zero portability.

Detailed analysis
Let’s take five different scenarios to understand how valuable each offering is, in that context:

On-premise application unchanged in the Cloud
This is characterized by multi-tiered legacy systems, or Java- and .Net-based applications.
If you choose Amazon, you could treat the machine instance as another server in the datacenter and do the necessary changes to configuration. With Google, however, you would need significant refactoring of application and data logic for your existing Java application. With Microsoft, you could re-factor data if your existing application is an ASP.NET. Else, your refactoring effort could be significant depending on the complexity of your application.

Web Application
Here, the assumption is moderate-to-high traffic with a back-end store and load balanced tiers.
Choosing Amazon gives you the same flexibility as the option above, but you would have to manually configured scalability and elasticity. With Google, you can use dynamically scalable features of AppEngine and scripting technologies to build rich applications.
Microsoft allows you to build scalable web applications using familiar .NET technologies. Scaling up or down is purely driven by configuration, aided by additional platform services such as Identity and Messaging.

Parallel processing, computational application
Here, we assume automated, long-running processes with little to no user interaction. Amazon requires you to configure multiple machine instances depending on the scale needed and then manage the environment. Features such as MapReduce allow for easier developing and deploying of compute-intensive applications. Google’s platform has minimal built-in support for building compute-heavy applications. Certain application scenarios such as image-manipulation are easier to develop with built-in platform features. Microsoft scores here. With worker roles and storage features such as Queues and Blobs, it is easy to build a compute-heavy application that can be managed and controlled for scalability and elasticity.

Application in the Cloud interacting with on-premise data
Applications in Amazon’s EC2 server cloud can easily be configured to interact with applications running on-premise. That’s a thumbs-up. However, Google offers no such support to allow for this scenario. You can render this possible through each application using an intermediary store to communicate with the on-premise application but that’s not a scenario Google App Engine promotes actively.Microsoft scores high again here. With features such as Service Bus to Sync platform components, it is possible to build compelling integration between two environments.