Why Storage Clouds Are Worth a Closer Look
It's time to take another look at remote storage services.
A few years ago, when application service providers (ASPs) were being hyped by IDC and others, I did a book-length dive into the ASP model. My conclusions at the time were that (1) ASPs were not about to become the dominant model for sourcing software as a service and (2) remote service providers, both for software and hardware platforms, were likely to go belly up. The bursting of the dot-com bubble helped put a fork in the nascent service delivery model, but not before prospective consumers did.
Customers wouldn't abide by the rules that were core to the business models of the service providers. Specifically, they wouldn't share infrastructure with their neighbors, preventing the realization of the "economies of scale" expected by providers from sharing the same servers, networks, and storage platforms among multiple clients. Customers who did use remote services almost universally demanded their own (unshared) platform, often because they were worried about their neighbor's data contaminating their own.
Other concerns that I had at the time (circa 2000) were that application server software had not evolved the robustness that consumers wanted, that cost-effective load balancing was lacking in most plays, and that security was a buzz kill for everyone involved. Time was needed for technology to catch up to the idea.
In addition, hardware vendors didn't exactly cotton to the storage service model. EMC and others descended on the likes of Storage Networks and other early providers with the time-honored goal of selling silos of spinning rust. The design of the Symmetrix and the pricing model it encouraged played havoc with both the service provider's "shared everything" architecture and with its value proposition: the cost of using a storage service turned out to be not a lot less than the cost of fielding and managing gear yourself. When the service provider's costs per TB were lower, the variance was unsustainable when customers demanded a "shared nothing" model to which they were accustomed in their own shops provided on premium-priced gear that often they themselves could not afford to field.
There were also instances of outright sabotage. Some IT administrators feared the change to an external service provider model as a threat to their own jobs. This fear manifested itself in everything from FUD mongering (raising fear, uncertainty, and doubt to drive unrealistic service-level guarantees) to heel dragging in the cutover phase of the agreement.
Bottom line: many folks lost a lot of money investing in ASPs and their cousin, storage service providers (SSPs). This explains why I have been reluctant to write anything about the latest craze: cloud computing.
A month or so ago, I hitched a ride from a hotel where I was speaking to the San Diego airport with a fellow who is a system architect with storage cloud provider Nirvanix. He chatted about his firm, its business goals, its "secret sauce" technology, and what changes he believed had occurred in the market that should encourage me to reconsider my views about cloud storage.
There had been successful storage service providers on the Web for several years, he noted, such as Salesforce.com, providing data protection for desktop users via the Internet. The downside of these services, I observed, was evidenced in the wake of Hurricane Katrina. Many on-the-Web backup service providers were hard pressed to return data to all of the customers who were impacted by the storm and wanted their backups returned ASAP. The queue was long and, from what I've been told, some folks waited as long as 100 days to get their data restored. He was not dissuaded.
The Nirvanix system architect said there were significant cultural and technological shifts since the 1990s. Secure socket layer (SSL), he said, had been much improved. Security is more sophisticated now and folks are more comfortable with the protection it provides.
As for the experience of the Web backup service users, he was quick to point out that Nirvanix is strictly B2B (business to business, I reminded myself about the old dot-com era jargon). Nirvanix provides archive, backup, and media storage services to big outfits that might use them internally or repackage them as services to others. Recently announced customer win General Electric is an example of the former, while Freedrive is an example of the latter -- a Nirvanix customer who repackages services to consumers.
He went on to contrast Nirvanix with other recognizable names in this much evolving space. Nirvanix calls itself a storage delivery network (SDN) and not a content delivery network (CDN) such as Akamai or Limelight. He says this makes them unique: "the first marriage of cloud storage to content management."
The Nirvanix system architect maintains that Nirvanix is also a different creature than Amazon S3. He says that the company originally saw Amazon as a competitor but quickly realized that the bookseller is "processor driven" and not "storage driven." Nirvanix' SDN provides Tier 2 NAS infrastructure to customers using storage nodes spread over seven data centers in six continents. The Nirvanix secret sauce is a single namespace provided via a combination of a virtual file system layer -- the Nirvanix Internet Media File System (IMFS) -- and inter-intra node load balancing.
Recent tests performed by Webmetrics and 3Tera claim that Nirvanix delivers 200 percent greater performance in downloads of large files than its closest competitor, Amazon S3. Plus, with the Nirvanix SDN, users can access their data from multiple locations simultaneously.
Ultimately, the success of Nirvanix might just come down to cost. Comparing the cost of homegrown Tier 2 NAS infrastructure and the cloud storage provided by Nirvanix, the numbers break down to $15 per year per GB for on site NAS storage versus $3 per year per GB for Nirvanix SDN. Arguably, data is far more secure and accessible in the Nirvanix SDN than on, say, a couple of NetApp Filers deployed in your own network.
I was impressed by the Nirvanix executive and see enormous potential for his cloud storage offering. Watch this space for further developments. If you are using cloud storage, and have an observation or comment to offer, contact me at firstname.lastname@example.org.