Jeff Bezos Chose The Wrong Forcing Function For AWS

Posted: Apr 10 2014, 5:35am CDT | by , in News | Technology News

Jeff Bezos Chose the Wrong Forcing Function for AWS
Photo Credit: Getty Images
/* Story Top Left 2010 300x250, created 7/15/10 */ google_ad_slot = "8340327155";

Amazon CEO Jeff Bezos loves forcing functions. At some point before 2010, Bezos decided that it was time for Amazon Web Services to make some rapid progress, so he started a project to move the computing infrastructure for on to AWS. As you would expect, AWS had to be adapted to handle, and became a stronger product with respect to handle web-scale applications. no doubt became a better and more efficient application in the process. This is just the result that Bezos was looking for.

While AWS made progress because of this forcing function, it is reasonable to ask: Was there a better forcing function? Recent moves by IBM and enterprise-class cloud firm Virtustream shows that Bezos may have made the wrong choice in forcing functions if he were seeking the largest market possible for AWS, not just increased operational efficiency for

Bezos had his eye on the wrong set of applications when he conceived his forcing function. My argument is quite simple:

  • Modern web-scale applications are built to run on the AWS cloud, which is constructed out of components that offer weak SLAs for performance.
  • A massive share of the $3.8 trillion that Gartner says will be spent on IT in 2014 goes to supporting traditional enterprise applications, which were built to run on components that are scalable and reliable.
  • These applications will be with us for decades and won’t easily be rewritten to use a web-scale cloud, but they could benefit from cloud hosting.
  • A better forcing function would have improved AWS with respect to traditional enterprise applications.

In other words, the forcing function Bezos should have pursued was one in which he approached 10 companies from a selection of different industries such as finance, manufacturing, retail, pharmaceuticals, and logistics and then asked them to work with Amazon to move as much as possible of their data centers to AWS.

The problem with this forcing function is that it may have driven Amazon back to the drawing board. The digital DNA of AWS was not designed for scalability and reliability of the individual components. Rather, it was designed to meet the promise of high levels of automation, rapid deployment and scalable capacity. AWS has made progress toward meeting the requirements of enterprise computing through dedicated instances and other means, but these are band aids, not the top-to-bottom engineering needed to really solve the problem.

“AWS started out without the notion of strong SLAs for performance. As they mature, they are bringing on more heavy duty solutions for performance of IOPS, computing, memory, etc.,” said Sean Hull, an IT consultant who specializes in high performance architectures and is author of the Scalable Startups blog. “Some enterprise apps will be able to run on AWS and some may not. In my experience, to get consistent high performance when using AWS, there are still many components such as databases that benefit from dedicated hardware.”

The World of the Polyglot Cloud

In the past five years, both IBM, Virtustream, and others have been working on the enterprise forcing function mentioned above, and they have made progress. It is worth looking at the difference between how these clouds work and what AWS offers. The bottom line is that we now are living in a world of a polyglot cloud. It is possible to move most of the apps in the existing data center to a cloud and run them there reliably without having to compromise on performance, security, or scalability, and without having to rewrite the applications. (See “How to Build a Reliable Cloud for Legacy Apps” for a longer discussion of app architecture.)

IBM and Virtustream have different strategies for constructing the enterprise cloud. Today we examine Virtustream and I will check in with IBM about its strategy in the near future. (My starting point for my look at IBM is this article (“How SoftLayer is Beating Amazon in Supporting the Application Lifecycle”) written just before SoftLayer was purchased by IBM.)

What Virtustream has done is broken the unit of computing down further so that it can solve the allocation of resources problem. Using an abstraction called a microVM, Virtustream can define a set of resources that has enough computing power, memory, IOps, and Network Ops to run an enterprise application. Unlike AWS, where you get an instance but no guarantee of performance, microVMs come with hard SLAs. This means that enterprise apps can be simply moved over to a Virtustream cloud, without modification or worries about degraded performance. Of course you can move any application you want to AWS, and if performance doesn’t matter, then there will be no impact. But if a person is waiting for a consistent speedy response from the app, you may be in trouble.

Like other clouds, Virtustream offers elastic capacity, so if you need a temporary increase in microVMs, you can get it. Unlike most other clouds, Virtustream has enterprise friendly features for compliance and security such as info about geo-tagging, automated compliance processes, risk-weighted threat analysis. In addition, Virtustream supports ESX, the hypervisor used by VMware, and KVM, the hypervisor used by OpenStack. This allows apps that run on VMware or OpenStack to be easily moved to the Virtustream cloud.

Virtustream’s cloud pricing is based on compute resources used, not compute resources allocated as with AWS and many other clouds.

The point of this discussion is that it is time for CIOs to start seriously considering a polyglot solution in which enterprise apps live in an enterprise cloud, web-scale apps live in a public cloud, and only the most high performance, mission critical apps stay in a data center.

How the Enterprise Cloud Market is Changing

I spoke recently with Rodney Rogers, CEO of Virtustream, about the increasing pace of enterprise cloud adoption and the forces driving it.

Rogers said after the company was founded in 2009, they had an uphill battle to convince buyers that they could really run their crown jewel applications on the cloud. But in the last half of 2012, thinking changed, and the early adopters and the first wave of early majority arrived in force, increasing the size of the sales pipeline by a factor of 10 in 6 months.

Rogers said that the fastest sales come from companies who have implemented virtualization but don’t feel they are getting a significant benefit. Customers in this category find they save from 50 to 60 percent of their budget for running an application by moving to Virtustream’s cloud. When companies have done a good job of implementing virtualization, Rogers said they can save from 20 to 30 percent. Moving from a bare-metal data center is the least likely move because of the increased automation required to adapt to the cloud. But companies that do manage to make this transition save from 60 to 70 percent of their app-related IT budget.

Rogers said that the following factors are also resulting in migrations of enterprise apps to the cloud:

  • Increased need for the cloud as an integration point. Enterprise apps often need to connect to big SaaS apps like or Workday or little SaaS apps like Box or Zoho. Making and managing such connections is easier from the cloud. Rogers said Enterprise apps once moved to the cloud are simpler to connect to SAAS applications – less security barriers to cross.
  • Increased pace of mainframe migration. Much of the computing workloads that used to be served by mainframes are increasingly able to run on the X86 platforms, which have become far more powerful. When such a migration is considered, it is possible to run apps on a beefy collection of microVMs.
  • When a hardware refresh is needed, apps increasingly end up on enterprise class clouds.
  • Companies migrate to enterprise class clouds to take advantage of increased security and ability to demonstrate compliance with various standards such as PCI.
  • Disaster recovery becomes far easier to implement and is more flexible in the cloud.

Rogers said that in addition to selling directly, Virtustream is finding a large number of customers through channel partners such as telecom companies in various emerging markets who are eager to offer an enterprise class cloud and a suite of surrounding services.

As Hull pointed out, Amazon is on the job with respect to closing the gap with respect to SLAs. Virtustream, IBM, and the other players are betting that the lead AWS has in building a robust ecosystem and broad functionality can be overcome by top-to-bottom engineering focused on what enterprises need.

Follow Dan Woods on Twitter:

Follow @danwoodsearly

Dan Woods is CTO and editor of CITO Research, a publication where early adopters find technology that matters. For more stories like this one visit

This story may contain affiliate links.


Find rare products online! Get the free Tracker App now.

Download the free Tracker app now to get in-stock alerts on Pomsies, Oculus Go, SNES Classic and more.

Latest News


The Author

Forbes is among the most trusted resources for the world's business and investment leaders, providing them the uncompromising commentary, concise analysis, relevant tools and real-time reporting they need to succeed at work, profit from investing and have fun with the rewards of winning.




comments powered by Disqus