Service Provider Models
The first thing you’ll need to figure out is what type of cloud offering you prefer.
Companies that want to manage their own storage and create private clouds should choose an Infrastructure as a Service (IaaS) vendor. In this model, the provider delivers hardware, bandwidth (if it’s a hosted private cloud) and the virtualization technology. End users must then provision and manage their own storage.
With the Platform as a Service (PaaS) model, the provider offers a storage platform, but does not delve into the specifics of data configuration settings and such.
Finally, if you choose a SaaS provider, or a public cloud storage provider, you will get a full backup and recovery application suite, and your data will reside in the cloud provider’s data center. At least some of the responsibility for data protection, availability, security and compliance is now handed off to the service provider.
The choice in models really boils down to two key decisions. First, how much storage proficiency do you have in your IT staff? If your IT team is already understaffed and overworked, the arrows point towards SaaS solutions. Second, how critical is data ownership to you? If you must maintain complete control over your data and don’t want to assume any of the risks involved with handing it off to third parties, it’s probably best to go with an IaaS vendor, so you can build a private cloud for your backups.
PaaS is sort of the Goldilocks solution. Not too private, not too public, and it will come with hooks to a specific vendor (like VMware or Azure). It can be a smart choice if you plan to deploy a hybrid cloud.
Software-Defined Storage (SDS) is a concept that’s gaining momentum with cloud storage providers. A snarky fellow on Wikipedia (and, no, I’m not being casually sexist; the editors on Wikipedia are nearly all men) argues that SDN is primarily a “marketing theme for promoting storage technologies.” That’s in the very first line of the entry. Clearly, SDS vendors are worried about issues other than editing Wikipedia entries.
There’s a kernel of truth in there, but SDS is actually a valid concept.
The SDS concept follows closely on the heels of the Software-Defined Networking (SDN) idea pioneered by startups Nicira (acquired by VMware for $1.26 billion), Vyatta (in the process of being acquired by Brocade), and Embrane.
Basically, SDS and SDN are familiar stories. It’s pretty much the virtualization and cloud story applied to networking and storage. The goal is to separate the management of the bits and bytes (or packets) from the underlying hardware. With SDN, the concept (and OpenFlow is a huge part of this) could help eliminate expensive routers and switches that rely on proprietary operating systems in favor of commodity hardware turned into networking boxes with open-source OSes.
With SDS, some analysts would place any virtualized infrastructure in the SDS camp. That’s probably an oversimplification. What SDS does that is different is that it turns storage into an extension of the hypervisor or operating system. Rather than having a dedicated VM or even an appliance with its own OS, you can basically turn all of your storage into essentially one big virtual hard drive.
In other words, with SDS you don’t need to worry about hardware and OS compatibility on the provider side. What does it matter? The storage has been abstracted from those issues. Of course, it’s never that simple, but that’s the vision — and really the roadmap — for where cloud storage is heading.
This doesn’t mean that startups positioning themselves as SDS vendors, such as Nexenta and ScaleIO, will be the eventual winners in the cloud backup space. Rather, as is so often the case with startups, it simply means they’re pointing the way. Winners are yet to be determined.
Architectural And Backend Technologies
If the cloud storage market is moving towards SDS, and I believe it is, then the backend technologies aren’t really the end user’s problem, are they? I strongly believe that eventually, most hardware will be commodity hardware, and your main choice will be among service providers (AWS, Rackspace, AT&T, etc.). If you are building your own private cloud, many of your choices will be predetermined by the IaaS vendor you go with (VMware, Eucalyptus or an OpenStack- or CloudStack-based provider).
One other thing to consider is the issue of remote versus local backups. Due to networking challenges, cloud storage is best suited to data protection, retention and archiving — for now. Disaster recovery from the cloud will be a slow, tedious process when compared to recovery from local disk or even from tape.
However, in this age of extreme weather, it’s a good idea to have a backup plan for your backup plan. So while recovery from the cloud is not ideal, it should be part of your disaster planning.
Industry Controversies And Debates
What will the future of storage look like?
Will the SDS model actually win out? Or will hardware make a comeback?
After all, there are some inherent risks in a software-first storage world. If hardware becomes too commoditized, will quality suffer to the point that you just can’t trust it? Sure, you can move data and VMs around at will, but will you need a carnival funhouse hall of mirrors of endless backups to ensure that crappy hardware isn’t your undoing?
For my money, the high-profile acquisitions of SDN vendors points to a software-first cloud world. (I mean, that’s really what cloud computing is about anyway, isn’t it.) And plenty of investment money is being funneled to SDS startups too.
Of course, legacy equipment and architectures tend to linger for years and often even end up having real staying power in specialized niche areas, but the storage market looks to be moving towards SDS.
Many experts cited security as the number one industry controversy. Fair enough, but I see this controversy as more of an obstacle, so please refer to the “Trust and security concerns” entry in the “Obstacles” section above.