The cloudy infrastructure bandwagon is so crammed full of server makers and cloud hosting providers that no one can see where it is heading. And so the box counters at IDC grabbed the reins this morning and gave them a good firm yank.
The market researchers have just put together a server cloud computing forecast, called the Worldwide Enterprise Server Cloud Computing 2010-2014, and as cooler heads have been saying for some time (and those whose heads are neither lost in the stratosphere nor up some warm, dark place), the advent of cloud computing is changing the server market, but not toppling it.
According to Katherine Broderick, a research analyst at IDC who follows enterprise platform and data center trends, IDC has a very formal definition of when a server is being used in a cloudy way. The server has to make use of converged networking for servers and storage, it has to be virtualized, and it has to be highly automated.
Cloudy infrastructure also has to have the means of keeping track of who is using what server resources and of hooking into chargeback systems so users can be billed, utility style, for the resources they consume running their applications. A cloudy server can be used in a public cloud, like Amazon's EC2, or in a private cloud, like a mainframe-based parallel sysplex cluster.
Yup. That's right. The $7.3bn that IDC estimates was spent on cloudy servers in 2009 had a very large mainframe component to it.
Broderick says that mainframes meet the IDC definition of a cloud, just like a stack of x64 servers running VMware's vSphere 4.0 with some other tools and using converged storage and network fabric would. Even VMware calls what it has created with vSphere a 21st century software mainframe, after all.
Yes, this is stretching things a bit, probably to the tune of $3bn to $4bn in mainframe server sales a year. But if a mainframe were not a cloud, then these cloudy server numbers would not sound so big. And Big Blue would get annoyed and maybe IDC would not have access to IBM information that it needs to do its estimating.
And so when you add the cloudy mainframe server sales to the cloudy x64 server sales and the smidgen of cloudy RISC/Itanium revenues, you come up with that $7.3bn for private clouds.
Broderick says that mainframes currently make up a big slice of private cloud server spending, but over time, x64 servers are expected to dominate. On public clouds, as you might expect, the x64 server is the platform of choice, and it will be highly unlikely to see that change.
IDC reckons that companies shelled out some $46.2bn on servers (at the vendor factory level) in 2009, and that means cloudy servers accounted for a total of $7.88bn, a 17.1 per cent share of the total server pie, including public and private clouds. Public clouds are a relatively miniscule part of the business, accounting for only $582m in sales compared to that $7.3bn for private clouds.
IDC is projecting that cloudy server revenues across public and private clouds will grow to $12.52bn by 2014, an increase of 59 per cent over those five years. Private cloud revenues will grow by 61.6 per cent over those five years to $11.8bn, but public clouds will only grow by 23.4 per cent, to account for $718m in server revenues.
You might have been thinking this number should be a lot larger. Like billions and billions of dollars. But Broderick says to meet the definition of cloud-related server revenue, it can't just be a virtualized server. It has to be automated and use converged infrastructure (networking and storage feeding into the server).
That growth for either private or public cloud spending on servers is going to be a lot better than the market at large, which IDC is projecting will only be 14 per cent bigger in 2014 than it was in 2009. In fact, if you take the cloudy server spending (both public and private) out of the server equation, then the over those five years the non-cloudy server revenues will only grow at a rate of 4.7 per cent.
IDC says that even though the virtualized server instances available through public clouds have lower average selling prices than real x64 servers, public clouds are going to be less broadly adopted than private clouds because companies are understandably jumpy about putting their ERP systems and databases out there on a cloud because of security concerns with a shared utility. It is hard enough to deal with securing the border between internal systems and the Internet; having to secure and prove the security of virtual server slices inside a public cloud is just too much of a hassle for IT shops.
And quite frankly, if this were not the case, IT managers and system admins would be voting themselves out of a job if they decided to use public clouds for mission-critical applications.
What Broderick says will take off on public clouds is email and collaboration for small and medium businesses that don't want to have their own infrastructure, as well as software applications like Salesforce.com and NetSuite that give end users easy access to slices of applications that are, for all intents and purposes, a shared cloud. (Server revenues among Amazon, NetSuite, Salesforce, and others are part of the public cloud server revenues mentioned above.)
Private clouds are going to do what systems in data centers currently do: run ERP systems and other decision-support applications as well as infrastructure workloads. They will just be a bit more moist than their physical and virtual server brethren. ®