Hybrid clouds 2012: the private cloud myth lives
Marketeer snake oil for scared CIOs
Hybrid clouds are all the rage in cloud computing today, with Gartner naming them "a major focus for 2012", even as hybrid clouds constitute fully 20 per cent of enterprise clouds today. But are they really anything more than a new face on private clouds? Marten Mickos, chief executive of private cloud company, Eucalyptus Systems, doesn't think so.
As Mickos pointed out to me over email, private clouds don't actually make sense unless they have a hybrid ability. Yes, he went on to say, there are and will always be some specific private clouds that are private and only private. But they are few and far between.
So why do we make the private/hybrid distinction?
Marketing. Early on, risk-averse CIOs needed to hear comforting words from the cloud computing industry that all their sensitive data wouldn't be hosted in an Amazon data center. Because, you know, data centres among other things deliver perfect uptime and are impervious to security breaches. Or maybe Not. So they were told that they could get all the benefits of cloud computing by building their own private clouds.
Which is, of course, utter and complete nonsense.
Many of the benefits of cloud computing go out the window the minute a CIO runs a cloud behind their firewall, as Amazon has cogently argued. And Amazon and other public cloud providers are arguably better positioned to invest in uptime and security than isolated enterprises.
For these and other reasons, CIOs are increasingly looking to hybrid clouds that blend the benefits of private clouds with public clouds. The ways in which this happens are myriad and include:
- Develop on public, deploy on private
- Develop on private, deploy on public
- Deploy first on public, then move workload to private
- Deploy first on private, then move workload to public
- Share workloads across clouds (cloudbursting)
- Run sensitive or data-intensive workloads on private, and other workloads on public
- Use one cloud as a back-up for the other
- Use private clouds as back-ups for other private clouds
At a certain point, the distinction between private/public/hybrid becomes unhelpful. With a common API - one claim to fame for Eucalyptus is its tight adherence to the Amazon API, which makes it easy to move workloads between Amazon Web Services and Eucalyptus private clouds - hybrid clouds become both easy and standard fare. Perhaps the distinction is necessary for CIOs anxious to keep track of all their cloudy assets, something that HP warns is harder and more critical in the cloud.
And, yes, there is an important technical difference between public and private clouds: public clouds are typically built on absolutely homogeneous infrastructure with - among other factors - just one type of CPU, one type of operating system and one type of hypervisor. Private clouds need to operate on top of highly heterogeneous infrastructure. For this reason, the software design must be different.
Still, as Mickos points out:
In a philosophical sense, the delineation between public and private cloud isn't that clear. Seen from planet Mars, Amazon Web Services looks like a private cloud. Red Cloud of Cornell University is a private cloud in many ways. But it is open to all researchers, so you could also say it's a public cloud.
As of today, there aren't that many hybrid clouds in production. But there are some, both from Eucalyptus and others like VMware. And this isn't surprising, since there aren't many publicly referenced private clouds in production, either.
But rest assured, when they do get announced, they're likely to have a hybrid component to them. Because that's exactly what private clouds are, or very soon will be: hybrid clouds. To adopt the cloud completely behind the firewall is something that might make sense in a white paper, but it's unlikely to drive the efficiencies and agility that increasingly drive cloud computing adoption. Hybrid is the new private. ®
Matt Asay is senior vice president of business development at Nodeable, offering systems management for managing and analyzing cloud-based data. He was formerly SVP of biz dev at HTML5 start-up Strobe and chief operating officer of Ubuntu commercial operation Canonical. With more than a decade spent in open source, Asay served as Alfresco's general manager for the Americas and vice president of business development, and he helped put Novell on its open source track. Asay is an emeritus board member of the Open Source Initiative (OSI). His column, Open...and Shut, appears three times a week on The Register.
"Run and owned by someone else"
is why my company won't touch public cloud computing with a 90 foot pole. We already have a remotely-accessible server onsite and two offsite backups where the sales and other field staff can get to their data while on the road / at home / wherever.
If we follow the sheep and start using public cloud facilities, how long would it be before the cloud provider or some other agency a) lost, deleted or tampered with the data; b) sold it to competitors; c) failed to implement security measures to prevent hacking or industrial espionage?
This endless push, push, push, to get our data out under control of other companies is really getting very tiresome. There are NO benefits a public cloud provides that can't be met by a private cloud setup, and a whole lot of things that can go wrong.
I keep wondering
how "cloud" is different to conventional shared hosting anyway, when we see well-publicised and wide-ranging outages from providers as large as Amazon and Microsoft.
I thought the whole point was that you didn't care about the implementation or provisioning or anything else, just that you get to run your stuff as needed. Then the cloud provider measures the required capacity, loads etc, with the cloud software automatically using the available hardware across a lot of different customers, so everyone benefits from the economies of scale such a flexible setup allows.
A 'private cloud' has to be a misnomer. At that point it's just a rack with VMotion...