Object Storage: Is it Soup Yet?
It may hold the key to getting a handle on unstructured data.
I know, I know. The reference in the title of this blog post, to the tag line from a 1970s Lipton Instant Soup commercial, dates me. But the memory came bubbling up from the deep storage in my brain as I listened to a couple of vendors recently describing their latest offerings in the object storage market.
Talk about time-worn references; "object storage" is a lot like "holographic storage," another technology that's been characterized as "just around the corner" since the Kennedy Administration. Truth be told, object storage has always had an appeal. Placing so-called unstructured data, a.k.a. user files, into a more robust and database-like configuration will provide a foundation for wrangling all of those bits into a much more orderly and efficient state. That, in turn, will avail files of real data management, which makes sense as a strategic way to bend the storage cost curve and improve information access.
What a Wonderful World -- of Data Storage
That's what object storage is really all about. Done properly, each file (or "binary object") can carry around a lot of highly granular metadata to describe its business context and content. That in turn provides handles on file-objects that allow them to be grouped in intelligent ways and provides a way to apply policies and services intelligently. Object storage, in short, would bring about a wonderful world of wonderfulness in data storage.
The sad truth is, however, that the market hasn't been keen to adopt it. A few years ago, the technology saw an uptick of interest when we became concerned about the impact of the Internet on file system storage. To recap: just as IPv6 was needed to increase the number of node addresses in order to keep pace with multiplying servers and workstations, some analysts argued that file systems would need to be replaced by object storage to cope with the explosion of files that seemed poised to outstrip the namespaces of most server file systems. Today, with improvements in file systems themselves, that concern about hitting the file system storage wall, and hence the incentive to consider object storage as a replacement for file system storage, has waned.
Still, a more fundamental problem remains. The growth of storage capacity demand, especially in virtualized server environments, has many firms worried. As heralded in the prognostications of IDC (projecting 300 percent year-over-year growth in storage capacity demand in highly virtualized environments) and Gartner (650 percent, per year), storage is becoming a huge cost center in IT.
Even with the prospect of storing files up on limited access Amazon Glacier storage at $0.02 per GB per year, companies bristle at the strategy as unsustainable. Ironically, at the hyperscale data center providers such as Amazon, Google, et al., interest in object storage as a means to economize on storage has never been greater.
For a time, de-duplication and compression technologies were touted by traditional file and block storage vendors as a fix for rising storage costs. However, most IT planners have learned that these technologies are merely tactical band-aides to the bigger problem of unmanaged data, which sees five to seven copies of the same file stored on disk and flash arrays. These files are seldom if ever accessed, consume electrical power on an on-going basis, and generate heat that taxes even the robust HVAC systems to their limits.
What's needed to address the storage problem strategically is a means to automate the management of data, from its initial high activity period through its long term, seldom-re-referenced-but-never-deleted state. A precursor to developing a strategic data management program is to place files into an object storage repository.
Promise Technology seems to get this. Their VSKY product offerings, which debuted at the NAB show in April, were developed initially to meet the needs for media and entertainment companies that were growing file-based repositories by Petabytes annually, according to Jason Pan, Senior Director of Product Marketing. He says that the company is now delivering VSKY appliances, constructed from their storage software and commodity hardware, that can scale in a modular way to provide a cost-effective object storage repository -- at least for archival or cold storage data (i.e., data re-referenced infrequently or never modified).
Pan says the company is working on a software-only VSKY offering that can be used with any hardware to provide object storage. Ideally, this infrastructure will provide common storage for any hypervisor workload rather than being dedicated to a specific hypervisor stack, as is the case with the current VSK appliance offerings.
Meanwhile, object storage software maker, Caringo, has announced FileFly for its object storage software environment, Swarm. FileFly creates a file system-like interface to object storage. It's intended to assuage the concerns of those who may be reluctant to abandon what they know (file systems and network attached storage devices) for new object storage models, says Vice President of Marketing Adrian Herrera.
FileFly is a very interesting development in object storage because it seeks to reduce the disruption associated in many planners' minds with the introduction of object storage. Caringo has gone to pains to work invisibly behind familiar storage hardware from NetApp or virtually any storage controlled by Windows.
Simply deploy an agent that works with Windows server file system or NetApp FPolicy, and you can begin configuring policies for migrating older data off of primary file storage into backend object storage. The names of the files (now objects) may be maintained in the file system registry if desired, but access requests will be brokered by Swarm, Caringo's object storage environment.
The whole thing plugs and plays with current interfaces, including ACLs and SMB 1 through 3. FileFly plugs directly into the file system instead of emulating it, as in the case of other gateway solutions, using Samba or Fuse.
A Game of Inches
With FileFly, you can also, optionally, attach better metadata to objects being stored in Swarm. That gets us a little closer to data management nirvana. And such sensible innovations may actually inch object storage closer to the mainstream of data storage technology.
Jon Toigo is a 30-year veteran of IT, and the Managing Partner of Toigo Partners International, an IT industry watchdog and consumer advocacy. He is also the chairman of the Data Management Institute, which focuses on the development of data management as a professional discipline. Toigo has written 15 books on business and IT and published more than 3,000 articles in the technology trade press. He is currently working on several book projects, including The Infrastruggle (for which this blog is named) which he is developing as a blook.