Sunday, June 10, 2012. New York City.  Tomorrow kicks off one of the largest Cloud computing events in the United States, the Cloud Expo.  My esteemed colleague, Bill Zack, will be covering patterns for Cloud Computing and kindly, the organizers have offered him a free pass to distribute to last minute attendees who would like to hear his talk, and also attend the show.

I’ve been following High Performance Computing, Grid Computing, Cloud Computing for many years, both in the Middle East and in America.  I have taken these observations and included them as part of Slalom New York’s Cloud Computing Strategy.  Here is a list of five opinions on Cloud Computing in New York City.

1. Amazon Web Services viewed as the top player for Cloud in NYC

2. IaaS is sometimes another name for old school hosting

3. Decades old Software Patterns may not be applicable in the Cloud

4. The line between IaaS and PaaS blurs every day

5. This opinion will be stated during the week of Cloud Expo

See previous: Decades old Software Patterns may not be applicable in the Cloud

The line between IaaS and PaaS blurs every day. Soon we will have to redefine the industry accepted categories for Cloud Computing (IaaS, PaaS, SaaS).

Historically, a strong delineation exists between Infrastructure as a Service (the bones and circulation) and Platform as a Service (the digestive tract and organs).

Amazon first introduced its Cloud platform as Infrastructure as a Services (IaaS) and now is creeping more and more towards Platform as a Service (PaaS). AWS Elastic Beanstalk is a great example of this.

Microsoft first introduced Azure as a Platform as a Service, and today is creeping more and more towards IaaS – the full redesign of the VM Role are a great example. Likely we will see more announcements later this month from Microsoft of expanded IaaS features in Azure.

With the introduction (and later acquisition) of Cloud Foundry by VMWare, these lines have further been blurred. VMWare has been around for decades. Cloud Foundry serves an elegant way to slide from a forklifted application to a full fledged Cloud designed application. VMware is truly an established and credible solution for providing the virtualized environment. The combination of open source platform with the Service Level Agreement support by VMWare makes Cloud Foundry a serious contender if not leader in the Cloud provider space.

At Slalom, we use Tier 3 for our hosting. Tier 3 provides the old school hosting mentioned earlier; however they also support CloudFoundry and this blurred world of IaaS/PaaS. To add more fuel to the fire, Tier 3 supports IronFoundry, which is .NET on top of CloudFoundry. Now you can take advantage of the strong marketplace of .NET skillsets, in addition to the benefits stated above.

Why do should a developer care? Number one and foremost – it makes your job a lot easier to migrate an on-premise application the cloud. Right now, it is a challenge to take a WinDNA/DCOM type application and move it into AWS or Azure. With VMWare, you can forklift your app into a virtualized environment. Next, you can use CloudFoundry / IronFoundry to modify and add PaaS capabilities. The beauty here is that you don’t have to rewrite your application fully to take advantage of the PaaS capabilities.

If the lines between PaaS/IaaS continue to blur, we’ll have to come up with another taxonomy to categorize different types of cloud service providers.

Full summary here: Five Opinions on Cloud Computing in New York City

See previous: Infrastructure as a Service sometimes is another name for old school hosting

At the upcoming Cloud Expo, Bill will cover design patterns for Cloud Computing, however, here is a a light analogy that I like to use to explain what we mean by “Decades old Patterns may not be applicable”.

In the past, performance was king. The faster your applications ran, the better the design. The better the architect, the better the developer.

In the Cloud world, performance is a commodity. With elastic cloud, it’s easy to expand and contract your application to run faster or relax against processing demand. It’s not even just easy – you don’t have to think about it.

Consider this scenario. Let’s say you have a giant image file of a high resolution photograph (assume it’s big, roughly 2 GB) and you have been asked to superimpose a Bend Sinister on top of it (think Ghost Buster’s Logo).

In the past, the fastest way to do this would be using vector based graphics. Take the whole image, toss it up into memory, and then use an algorithm to draw the Bend Sinister on top of the image. Super fast, very performant – you get lots of pats on the back.

Today if you did that in the Cloud, you’d get some frowns. Compute time costs money. Nothing is free in the Cloud. Even storage costs money, and storage is cheaper than compute.

So in the Cloud, this same requirement for a Bend Sinister drawn on top of a photograph, may be best designed by creating two images – one of the original photograph and the other of a Bend Sinister. This keeps the drawing in storage, rather than compute. Cheaper. Then superimpose them onto one another.

There are many scenarios were traditional software design may need to be revisited to maximize cost and efficiency in the Cloud. Usually clever performance design techniques are the first place to start.

See next:

Full summary here: Five Opinions on Cloud Computing in New York City