Even if it didn’t bring mass-mainstream adoption to cloud computing, 2010 was still rightfully advertised the “Year of the Cloud.” The landscape really began to take shape, with cloud providers honing their offerings, important issues surfacing to the light of day and industry consolidation finally beginning to happen. In most areas, however, the action is likely just getting started, and we’ll continue to see the real picture of cloud computing emerge throughout 2011. Here are just a few trends we’ll be watching in the coming months.
More specialized clouds and cloud services: After seeing the response to Amazon Web Services’ specialized services such as Cluster Compute Instances and Elastic MapReduce, Heroku’s Facebook App Package and Joyent’s gaming-targeted offerings, other cloud providers will roll out their own specialized services. Whether it’s Rackspace, GoGrid or even Microsoft, there is plenty of opportunity to target these same markets, or perhaps go after new ones, such as mobile-app developers. Also, more specialized cloud providers should start popping up, like they’ve done in areas like rendering and graphical processing. It’s becoming pretty clear that there’s no room at the inn for more large-scale cloud providers, and most IaaS providers are left sharing AWS’s scraps. Those wanting to expand or start their cloud businesses will have to find a target audience, be it specific developers or vertical markets. The thing is, the cloud economics are still very appealing — no one can argue with the prospect of on-demand resources and slashed capital and operational expenses — and there is still a large untapped pool of potential customers. Providers just need to find ways to reach those customers on their terms.
Infrastructure startups will get bought: Although there are myriad M&A possibilities in cloud computing and Big Data, three names, in particular, come to mind when thinking about 2011: Cloudera, Engine Yard and Eucalyptus. The first two are doing such good business in selling hot technologies — Hadoop and PaaS, respectively — to enterprise customers that large vendors needing to establish their strategies in these areas will come calling. Engine Yard has already been a rumored target for VMware, and any lingering interest might explain why VMware allegedly passed on Makara (which Red Hat bought in November). Even VMware doesn’t come calling, someone wanting an established customer base and a product that has room for growth beyond its Ruby roots has to make a compelling offer — especially now that Heroku is off the market.
Cloudera is the face of commercial Hadoop, and has partnerships across the database, data warehouse and business intelligence sectors. Who else would a large vendor trying to establish its Big Data strategy look at? Because of Hadoop’s complementary nature to existing data systems, EMC and Oracle seem like potential suitors, although Cloudera’s support for open source might lead it to be picky about where it ends up. Dell could be a dark horse, too, as Hadoop clusters seem a natural addition to Dell’s lineup of analytics appliances and prepackaged internal clouds.
Eucalyptus presents a different situation: It needs to either make a big move or get absorbed. The company deserves all the credit in the world for making on-premise, Amazon EC2-like IaaS a reality, but it appears in danger of becoming obsolete as a standalone company. OpenStack sucked much of the open source wind from the room, and hungry startups like Cloud.com and Nimbula are selling internal IaaS products with the advantage of having learned from both Eucalyptus and public cloud providers. There’s certainly value in the Eucalyptus technology, but the open-core, partner-centric business model just doesn’t seem sustainable against cloud-in-a-box offerings or those pushed by major vendors like VMware. Eucalyptus could certainly find a home within a company like Red Hat, or buy its way into a more-holistic offering.
The fusion of IaaS and PaaS will continue: PaaS is the future of cloud computing, and true-blue cloud providers that want to remain relevant in the long term will have to develop offerings of that ilk. Joyent is offering both already, and it wouldn’t be surprising to see AWS and/or Rackspace launch their own, too. If nothing else, offering PaaS will help cloud providers keep as much money as possible in-house by reducing the need for developers to look in different places for different applications. Interestingly, this is one area where software vendors are leading cloud providers, as VMware and Red Hat are already doing this for internal clouds. AWS, however, will have to tread lightly, as it already hosts a number of PaaS services atop its EC2 infrastructure, and it doesn’t want to burn any bridges or lose that hosting revenue. For proof that IaaS is still very much the cloud type of the present, however, look no further than Microsoft adding IaaS to its Windows Azure PaaS offering. Microsoft has one of the most-acclaimed PaaS offerings around, but customers and partners pushed Microsoft to add IaaS functionality so they could leverage the cloud for existing applications not designed for PaaS environments.
Low-power processors will have their day: With Tilera piling up server partners for its many-core RISC processors, Calxeda promising ARM-based servers and SeaMicro loading boxes with Intel Atom processors, something has to give in 2011 as it relates to high-powered x86 processors’ data center dominance. It’s far too early for even all of the alternative-architecture startups combined to make a dent in said dominance, but it only takes a spark to start a fire. Microsoft and other large data center operators have been public about their experiments with low-power processors, so there’s definitely an appetite for these types of servers. If Microsoft, Facebook, Twitter or any other large web property were to come out publicly as running even a small non-x86 server farm, it could spark a trend. These types of companies, which operate webscale data centers, know better than anybody how to squeeze costs out of better operations. That’s a capability that’s only getting more important.
A cloud provider will end up in court: We can all thank AWS for exposing us to some legal dark sides of cloud computing during 2010 (i.e., the Eli Lilly and Wikileaks issues), but those will just be appetizers if a cloud provider suffers a major data breach or loss. I’ve written in the past that cloud providers shouldn’t necessarily be liable for issues like outages or breaches, but that doesn’t mean they can’t be sued for them. Maybe the result will be that certain contractual provisions are found invalid, or maybe it will set some precedent as to what’s required from a security perspective. Maybe nothing will change. It’s little more than a gut feeling that such an occurrence will happen within the next year, but for a delivery model that is so steeped in FUD, it has just been too smooth a ride thus far.
