Cloud computing, still in its infancy but growing steadily as a major player in midmarket enterprise IT, is already starting to branch off into more custom-designed configurations.
Cloud, or utility, computing serves up computing power, data storage or applications from one data center location over a grid to thousands or millions of users on a subscription basis. This general kind of cloud — for example, services provided online by Amazon EC2, Google Apps, and Salesforce.com — is known as a “public” cloud, because any business or individual can subscribe.
Private cloud computing is a different take on the mainstream version, in that smaller, cloudlike IT systems within a firewall offer similar services, but to a closed internal network. This network may include corporate or division offices, other companies that are also business partners, raw-material suppliers, resellers, production-chain entities and other organizations intimately connected with a corporate mothership.
Obviously, security is much tighter in a private cloud, simply due to its exclusive nature. And security is Reason No. 1 that an enterprise, if sold on the cloud computing concept to begin with, will consider constructing its own private cloud formation.
How Dell approaches private cloud structures
Paul Bell, president of Dell Americas, told eWEEK Nov. 4 at a roundtable discussion at the Dreamforce conference in San Francisco that private cloud computing is indeed on the company’s agenda and that it is a trend in which customers are beginning to invest.
“This [private cloud enterprise] is definitely on our radar. We work with partners like Rackspace, which has been supporting smaller businesses, trying to leverage hosting that way. We are also seeing a growing number of customers trying to built it [their own cloud systems] themselves,” Bell said.
As a mainstream supplier of the hardware to put these things together, Dell is coming at this in several ways, Bell said.
“One situation is where we may have a joint [private cloud computing] project with Citrix or VMware — I think right now we’re doing about a third of VMware’s [hardware] business — in the server virtualization market,” Bell said.
“Increasingly, our customers are working on technologies such as VDI [virtual desktop infrastructure],” Bell said. “This allows them to do client virtualization. The reality is, there are multiple means to the end state today.”
VDI is a server-centric computing model that borrows from the traditional thin-client model but is designed to give system administrators and end users the best of both worlds: the ability to host and centrally manage desktop virtual machines in the data center while giving end users a full PC desktop experience.
“This reminds us an awful lot of where server virtualization was about three or four years ago. People were testing it, thinking about it, but it really wasn’t out of production at that time. Now it’s gone massively into the mainstream at a very rapid pace,” Bell said.
VDI for use in private cloud computing appears to be headed in the same direction. However, the different client virtualization technologies that are available “are still in kind of a fragmented state,” Bell said.
“There are many ways to do it. So we’re in a process right now of scaling out our consulting capability … to help customers decide upon how they would do that cloud,” Bell said. “There are a lot of issues, starting with, ‘What is the server backend?’ to networking issues to ‘What kind of device the client ought to be interfacing with?’, and so on.
“No question that for security reasons, this [private cloud computing] is a very strong trend for our customers.”
Private Storage Clouds All About Control
Storage Networks and SANRISE (acquired by Hitachi) were two early on-demand storage companies, among others, that got upended in the 2000-2003 bubble.
“These were about setting up storage structures at places like [Internet service provider] Exodus,” Sajai Krishnan, CEO of cloud storage software startup ParaScale and a storage industry veteran, told me. “Typically, they provided big EMC-type storage and sliced it up to offer to enterprise-type customers. That was the ‘Gen 1’ of cloud storage.”
It didn’t work because “nobody wanted to buy Tier 1 data storage in this kind of ‘cloud’ context, although it wasn’t called ‘cloud’ back then,” Krishnan said.
Online storage services have evolved a great deal since then. Companies such as EMC’s Mozy, Carbonite, Box.net, Amazon S3, CommVault, Seagate’s eVault, Iron Mountain Digital and others are primarily competing for consumer backup business.
On the business side, companies such as Nirvanix, Cleversafe and ParaScale (which is still in beta testing), have staked out their territories and are becoming the de facto, go-to folks for Web 2.0 companies looking to build out their own private clouds.
“The reasons to build a private cloud are very simple: They’re all about control and security,” Krishnan said. “‘I want it within my firewall; I want my people to manage it.”
Amazon’s Simple Storage Service (S3) online storage service is a very well-engineered service, Krishnan said, but “it operates at such a scale where sometimes you have these hiccups,” he said. “If you’re in S3’s IT shop, you have no way of actually solving that problem in any manageable time frame. And if you were to own your own cloud, you know exactly what’s going on.
“You know which software revs you have rolled out, you know which servers are choking up — all of this is possible when you know your constraints,” Krishnan said.
When does it make sense for an enterprise IT shop to think about building its own cloud?
“I believe that if you have 20 to 30 terabytes of data growing at about 10 to 20 TBs per year, then you should think about having your own cloud,” Krishnan said. “That’s about a minimum level. If you’re much smaller than that, and your growth is not much more than that, there are plenty of appliances you can buy with 12 and 24TB of capacity that will work just fine.”
Software for Cloud Storage: Not Easy to Do Well
“It’s hard to do the software for these storage clouds,” Henry Baltazar, storage analyst for The 451 Group, told me. “There’s two parts to it: The first part is the front end, the application. The thing that’s host-facing: How do you manage it? How do you provision it?
“The back end is the nuts-and-bolts part. How do you scale the thing out? How do you have enough buckets for all the storage? That’s where you’ll see that the new [second-generation cloud] companies like ParaScale, Nirvanex and Cleversafe know how to do this well,” Baltazar said.
High-performance, dedicated storage systems are typically used in government laboratories, such as Lawrence Livermore, Sandia and Los Alamos, which require high scalability and ultra-high throughput performance for high-end scientific computing projects. In a way, they resemble custom-designed cloud systems, but “they’re really more like big machines,” Baltazar said.
However, the newer companies noted above are providing infrastructure for unstructured data that doesn’t need particularly fast I/O access. This business data doesn’t need high security and simply needs to be a safely accessible place; small, starter-type clouds — namely, those consisting of two to five servers — can cost in the $10,000-$15,000 price range and are well within the budget limits of many mid-size businesses.
“The ParaScale model is designed to scale up inexpensively for a lot of backup and archiving uses, for that content that won’t be needed too often. It’s been said that 70 percent of stored data is never touched again,” Baltazar said. “That’s what you want to put in a storage cloud.”
Private clouds are a viable way to save business data in view of the recently amended Federal Rules of Civil Procedure, which require businesses that keep digital data to have a systematic process by which they can show the court, in the event of litigation, that it is keeping it organized and available for at least three years.
What does the future portend for private clouds?
“It looks good,” Baltazar said. “There’s plenty of room for this business model to work well.
“There’s tons of content, especially in the consumer space, that’s being created every day, whether it’s video, photos, office documents — that is not in Exchange servers, SQL servers, or Oracle servers. You don’t need to have high I/O for this, you just need to get to it.”