Amazon AWS really wasn’t designed for the Enterprise


Amazon Web Services (AWS) is “The” power house in Cloud computing. However, Amazon has struggled to gain a foothold in the enterprise. This is the case even when the geekiest of the enterprise SMEs finds creative uses for the service itself. EMC’s Chad Sakac provides a use case where his team builds an AWS application to test their ScaleIO solution for a total cost of only $7.41! I can see a C-level executive making the statement, “If EMC can find a use for AWS surely we can migrate our enterprise data center to AWS.”

After all, Netflix runs almost all of its streaming service on AWS and Netflix accounts for 1/3 of all Internet traffic at peak. If AWS can handle Netflix, then it can obviously support running a couple of Exchange and File Servers? AWS even offers a VDI solution to keep the users close to the data just like your expensive in-house VDI solution. I can understand the temptation to look toward a provider like Amazon to migrate your data center services.

The key to Chad’s example and the Netflix use case – both were purpose built for the AWS consumption model. Netflix is the undisputed master of leveraging cloud to run an IT operation. Amazon has built several tools that they’ve since open sourced that allow other organizations to take advantage of AWS like services. Netflix’s generosity reveals two of the greatest challenges for migrating traditional operations to AWS. The first is management of the cloud itself.

Enterprise service management solutions are all built around managing onsite solutions. These systems include applications such as Exchange servers, Windows servers, and Oracle databases. Retrofitting these operations and platforms for AWS is no small task. Netflix has practically built custom tools for service management.

More importantly, the challenge is the cost model for Infrastructure as a Service (IaaS) solutions such as AWS versus the use model of typical enterprise applications. EMC’s use case is a perfect example of a typical AWS workload. The EMC team had a specific task that was going to take a finite amount of time. EMC sized the AWS environment for 2 or 3 different tests and disposed of the workloads. Their experiment costs $8.00. Chad’s team leveraged super computer type resources for basically the cost of electricity to run a similar environment in a traditional operation. However, if you take a typical application such as Exchange that never shuts down then the costs are on the opposite side of the spectrum. Without getting into the math, you could pay for the server and operations of a traditional environment with just a few AWS workloads running a fraction of your enterprise applications.

Netflix is a prototypical Internet startup. Netflix built all of their operations and applications based on cloud architecture. Their apps scale the number of AWS workloads up and down as usage dictates. Netflix is able to limit their costs during non-peak streaming hours. During peak workloads, Netflix is able to scale as much as needed during the sprints of heavy usage. Enterprise applications do not behave in this fashion. In the traditional enterprise, when the application server becomes overwhelmed you buy a bigger application server. The applications don’t scale “horizontally.”

When looking at migrating to AWS it’s not a simple question of can your enterprise application run on AWS. But a question of how do you build new applications to take advantage of the cost and operating model of services like AWS. Alternatively, you could look at services such as VMware’s vCloud Hybrid Cloud Service (vCHS). vCHS consumption model is much closer to typical enterprise virtual machine environments. But it’s not a slam dunk.  The cost of vCHS and the like can be just as expensive. So, when presented with the question, “Why not AWS?”, the answer is sure.  The catch?  An organization has to be ready for a complete transformation of their IT operations and services.

The enterprise view of AWS and Cloud computing in general is about the transformation of IT operations and services.

Published by Keith Townsend

Now I'm @CTOAdvisor

9 thoughts on “Amazon AWS really wasn’t designed for the Enterprise

  1. Thank you for sharing. I also think the cost is _much_ cheaper to host it yourself (both AWS and vCHS). I use 500 VM on 40 ESXi hosts for 3 years as model.
    Full disclosure: I work for VMware as Account SE in Singapore.

  2. AWS application ? Cloud application, I would say.

    On Twitter it was hard to communicate what I was gonna say. I was away from a real keyboard, I tried to be short and to the point. 🙂

    I’ll try and explain it again here. I would summarize it this way:
    Cloud is especially useful for workloads that have a lot of peak demand. Netflix is a very good example. But you need to architect the application to be able to grow and shrink. Legacy application usually don’t fit that model.

    You mentioned Exchange, well I’m certain Gmail does grow and shrink depending on demand. I think it can be done. Changing the way Exchange works will take a lot of time. In the meantime there is an opportunity for an other company to get into that market.

    I think you could say: cloud is like solar power, self hosting is like the base load power plant.

    But that doesn’t mean ‘bursting to the cloud’ is a solution in many cases, because of the data. That data has to be replicated and kept in sync and prevent conflicts. That is an even harder problem for most applications than just handing growing and shrinking. So it is much easier to choose 1 place to run your application than 2.

    So you’ll have to choose on a per application/dataset basis.

  3. It’s too true.. I see DaaS and the rumoring of a full on private cloud soltuion as steps that Amazon is taking to accept the notion that the enterprise isn’t going to ‘come around.’ The good news is that the born in the cloud business is still growing hella fast.

  4. Good post. In my conversations with enterprises I’ve found that even though many are still keeping their existing production applications on-premise (on VMware/KVM etc) for exactly the reasons you outlined, they are looking to use the public cloud for bursty requirements like dev & test environments for those workloads. It makes perfect economic sense but begs the question: given that the datacenter and AWS are such different environments, is it possible to use the cloud for dev & test of on-premise environments? New technologies like nested virtualization are addressing this by encapsulating the entire application and allowing VMware workloads to run unmodified on AWS. See http://www.ravellosystems.com/blog/the-future-of-virtualization/

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: