Amazon’s Elastic Compute Cloud (EC2) is orders of magnitude bigger than its next largest Infrastructure as a Service (IaaS) competitor. At first glance, this would seem to imply that Amazon’s massive scale should give AWS a significant cost advantage over fledgling IaaS cloud offerings. But the […]

Amazon’s Elastic Compute Cloud (EC2) is orders of magnitude bigger than its next largest Infrastructure as a Service (IaaS) competitor. At first glance, this would seem to imply that Amazon’s massive scale should give AWS a significant cost advantage over fledgling IaaS cloud offerings. But the advantage Amazon gains from its scale is not necessarily on the cost side. Instead, it is in the growing list of startups racing to fill the holes in its Amazon Web Services (AWS) offering. But is Amazon eyeing the same product road map? To help answer this question, startups can look to VMware to understand the existential risk inherent in being a vendor in AWS’ ecosystem.

AWS is not without its faults. Notably, a person using AWS services like EC2 must still perform all the manual “plumbing” such as applying patches, managing configurations, etc, that they would have to if their servers were in their captive data center.  For entrepreneurs and investors alike, this creates an exciting opportunity: The sheer size of EC2 means that thousands of EC2 customers need a new toolkit to help manage their infrastructures in the cloud. This is also great for Amazon. While Rackspace needs to dedicate development resources to build a management console for Mosso, vendors like Rightscale are doing this work for Amazon. As Amazon’s ecosystem matures, AWS only stands to benefit.

In many ways, this parallels what happened in the virtualization management space. As VMware’s hypervisor, ESX, penetrated the data center, IT administrators realized their existing data center management tools were ill suited for virtual environments. So startups emerged to fill the holes in VMware’s only management console, Virtual Center.

But once VMware realized that the hypervisor was getting commoditized, it rushed to build a deep management layer toolkit that it could then sell with juicy margins. Just by announcing its vCloud product roadmap at VMWorld 2008, VMware paralyzed swaths of startups in its vendor ecosystem that had been trying to solve the various management pain points companies were feeling with their VMware implementations.

Amazon Web Services is almost by definition a commodity service. Consequently, Amazon, like VMware, is probably eyeing the management layer as an opportunity to generate high-margin revenues. This is not to say that there isn’t an opportunity to build a significant company by supporting AWS; Amazon can’t fight all fires in all customer segments as well as a specialized startup can. But Amazon’s entrance into the management space is probably less a question of if than when. Startups looking to innovate in the space would be do well to look at which startups have thrived in the virtualization management ecosystem — and which ones haven’t.

Sarah Tavel is an associate with Bessemer Venture Partners in New York.

  1. Having to apply patches and do your own configuration is not a bug, it’s a feature.

    Amazon doesn’t as a rule chase “high-margin businesses,” and they’ve already ceded the management layer to companies that operate in their ecosystem like RightScale.

  2. Your observation is a good one, Sarah.

    In terms of parallels with VMWare, to me, they seem still focused inwards on the compute-only part of the data center – whether it comes to security, policy, workflow orchestration, performance management etc. For security they have VMSafe for which a lot of work has to be done by an incumbent security vendor; and they have vShield which is only for corralling VMs within their compute environment. For a data center to run, the concept of vShield needs to span the other elements in the DC; even vCenter has a web services API available for others to exercise. This is the reason system mgmt incumbents like IBM and HP will continue to be the aggregators and majority margins will continue to go to them (See WSJ article on BestBuy as another non-virt, non-cloudy example – http://online.wsj.com/article/SB124078866665357525.html).

    Same goes for what VMWare seems to have done with B-Hive (by taking it from physical & virtual to virtual-only), same with Dunes Tech for the Orchestrator which again is for compute only tasks.

    I believe that when AWS does make the workflows (the ones you alluded to in the post) more automated, several opportunities will continue to crop for, say, tying the compute containers to the rest of the infrastructure – both intra AWS and inter-public and private ‘clouds’.

  3. I think Amazon is the opposite to Microsoft. They only want to develop what is necessary and happy to use as much open source as possible. Customers do require some management etc and some startup companies will have developed products that Amazon will replace. They are the Wal-Mart of computing. They are only interested in the high-volume, highly automated business. So they will not go into consulting or into specialistic areas which they cannot make high volume. In the longer term amazon may buy Ecalyptus so they can supply hybrid clouds which wolud move them more towards the enterprise software business.

  4. pool party Sunday, May 3, 2009

    There are opensource projects also trying to address the cloud management issues. Pool party is an instance http://github.com/auser/poolparty/tree/master . If things pan amazon may build of the available opensource tools :-).

  5. Sujit Mohanty Sunday, May 3, 2009

    If Amazon were to make a corporate acquisition, they would purchase Elastra. They funded part of Elastra’s Series B for $8M. Most likely to do an acquisition for down the road.

  6. Paul Lopez Sunday, May 3, 2009

    Sarah makes a good connection between the two. It depends if AMZN wants to be more vertically integrated. Sometimes we go with “Best of Need versus Best of Breed.” my quote, feel free to use :)


  7. in some ways elastic map-reduce is a step in this direction. people could already run map-reduce on ec2 and s3 – EMR bundles it all up with a management console etc. to make life simpler for lesser geeks.

  8. New Amazon AWS Public Beta Announced Monday, May 18, 2009

    [...] What Startups in Amazon’s Ecosystem Should Learn From VMware (gigaom.com) [...]

  9. @TheatroSaoPedro recebe a exposição fotográfica “Amazônia, Tão Perto, Tão Longe”, de Dulce Helfer, nesta terça! http://t.co/RMUkIEvD

Comments have been disabled for this post