Last year saw the long-expected, and oft-delayed announcement by Google that it was entering the cloud infrastructure market. Google Compute Engine (GCE) is a fairly immature (in terms of product breadth) raw infrastructure service that initially at least, seems to be trying to compete with the incumbent, Amazon Web Services, on price. A few weeks ago I got an email from Floyd Strimling, Technical Evangelist at Zenoss that invited me to have a chat about what he sees as the likely challenger to AWS for public cloud supremacy.
It was interesting talking to Strimling, given his perspective that OpenStack is being given too much attention as the competitor-in-waiting to AWS. It’s a natural enough perspective to point to OpenStack as having the biggest opportunity – with a plethora of companies signing up to the initiative (the likes of HP, Dell, Rackspace, VMware, EMC etc etc), it would, at first blush, appear to have the sort of traction that Linux was blessed with a decade or so ago. But Strimling sees it otherwise. He believes that Google Compute Engine will emerge as the clear alternative to AWS, in part due to the:
complex mix of datacenters, virtualization, breakthrough software, a comprehensive fiber backbone, and the use of new technologies that are crucial to Google’s own core business and the public cloud. The challenge with the hoisters jumping into cloud is they have no skin in the game nor any learning from years of struggling to make the infrastructure work.
On particular Strimling points to SDN as Google’s secret weapon. He points out that Google has been doing pieces of SDN for years while its relatively new for the OpenStack community. Where Strimling’s perspective gets interesting however is the resultant friction due to the purchase of Nicira by VMware. He posits that the acquisition, regardless of VMware’s entering the OpenStack community, is a cause for disquietude from both within OpenStack, and outside it.
Outside of Google’s undeniable expertise in the technical aspects of delivering mass scale infrastructure however, I fault Strimling’s perspective somewhat because of Google’s relative inexperience when it comes to selling and supporting enterprise-level products and services. Clearly Google Apps has had some success with enterprises, but there are also some glaring examples of Google’s promises not being met when it comes to product delivery. Contrast this with members of the OpenStack initiative, for all it’s difficulties of the moment, HP is clearly a company with massive enterprise credibility, as are Dell and VMware. Rackspace (disclosure – last year Rackspace sponsored my curation of the CloudU program) is a newer kid on the block but it too is differentiating itself of enterprise-level support.
It’s interesting to see the recently announced enterprise level support that Google is offering for GCE. It’s also interesting that third-party vendors like Cloudscaling are adding support for the GCE APIs – a vote of confidence that will be appreciated no end in Mountain View. All this is just tentative first-steps however – It seems to me that for Google to really hit it out of the park with GCE, it would take them becoming a far more open technology vendor, with a high level of communication between itself and its customers. Sure they can match on price and, over time, on the breadth of the service catalog, but they need to improve their fundamental approach towards the enterprise – they need talented accounts teams with technical evangelists (case in point – Werner Vogels from AWS who flies the flag impeccably), customer service teams, better and more clear licensing and privacy policies. Of all vendors, Google is most likely to be able to do it – they have the scale and they have the existing revenue streams if they chose to – the question is how much attention will GCE get in the face of the epic struggle within the ‘plex to get Google+ firing on all cylinders?
Either way – by the end of 2013 the cloud infrastructure space will be looking very different from what it does today.