Tuesday, March 17, 2009

Cloud Computing: Defending the undefinable

A panel discussion featuring representatives from Google App Engine, Amazon's Chief Technology Officer, and Microsoft.

Web Services provided by each company: 

Amazon has been offering cloud-based services like queing and db services, infrastructure and storage for a couple of years.

Google's App Engine is a scalable option that focusses specifically on web applications...versus some cloud computing services that include non-web computing options. Google's services scale automatically with your need, it doesn't require any action on your part. 

Windows Aja: An operating system for the cloud that includes services like storage, db (SQL) and tools for a rich programming environment.

Q+A

App Engine is built to work across multiple languages and regions. Amazon is currently available only in North America.

Can Windows XP apps run in the cloud? 
All apps must be written cloud compliant...standalone web apps in asp or .net are fairly straightforward.

Do your use your own cloud services? 
Amazon built its web services for itself...and that was the basis for creating an offering for customers.  Amazon is hosted on Amazon Web Services platform.

Google has some applications on app engine (Google Moderator, iGoogle for example). Many of hte smaller projects are launched via app engine. Using App Engine you are running in the same data centers as Google.  Some things in app engine are restricted to ensure scalability...this is having an opinion about how things should be built. You have to be prepared for scalability and app engine provides restrictions to serve as guidelines...stateless apps are an example. Databases with join.

Microsoft developed services to support their own use...many small teams developing applications. This long tail view of project development is what drove the need for a common infrastructure. 

How Will Cloud Security be managed? Will their be licenses for corporate use outside  the public cloud?

Amazon has no plans to launch a corporate label...but the tools are developed to help different verticals develop and administer apps to conform with specific industry requirements (such as HIPPA in healthcare)

Google has worked hard to ensure privacy and security in its data centers, but nothing specific about industry requirements.

Microsoft understands that different countries and industries have varying requirments. The Public cloud is difficult to manage across these variations. But private clouds will likely be the trend implemented to address these variances.

What Trends/Challenges Do You see in Cloud Computing?

Amazon: customers move from babysitting applications to increasing automation and horizontal scaling.

Google: Moving to the cloud changes how you build applications. Developing for distributed systems requires an understanding of details like latency and consistency that aren't as iportant for an app running on one server with one database. In some ways, this changes how programming gets done.

Microsoft: Automation gets built in. Its the biggest pain point for many.

What plans for Open Standard Support (to avoid vendor locking)?

Amazon: We constructed the services to provide tools, not frameworks. In essence, your apps can then run in other clouds. Standardizing the cloud will occur, but for now, there are only a few cloud providers, so customers must drive what standards need to be deployed.

Google: As a web company, Google thinks that moving in and out of app engine is good for everyone. The SDK will evolve to incorporate flexibility, but for now standards aren;t evident...if they emerge, Google wants to enable them

Microsoft: Lock in occurs at multiple layers...programming, data, etc. MS plan is to enable you to use your favorite stack. In the long term, private clouds may also influence standards. But data lock in is something noone will want to be held to.

How green are the clouds? (Not just hardware, but software too.)

Amazon: Efficiency + Cost effectiveness often go hand in hand with 'green' Because we are about driving costs and efficiency in our data centers we are pursuing this. But there are no standard metrics for measuring green code that each provider can publish and be held accountable to.

Google: Best thing that is comparable now is embedded in payments. As costs go down, they reflect a reduced cost of doing things (i.e., power, hardware, etc.) In general though, systems of scale are more efficient and provide opportunities for optimization that multiple approaches don't.

Microsoft: Costs of running a dataa center are predominantly driven by power and cooling. These costs are directly reflected in the cost to end users. Design of data centers,e tc., is continually evolving to drive out these costs...by reducing the power and colling requirements.

END





No comments:

Post a Comment