The tickler I received in email this AM filled me with interest in what Mr. Foskett, lately of Nirvanix, was going to write about this question in his multi-part blog. My fingers tingled with expectation as I keyed in the URL for his postings.
Alas, what I discovered disappointed me. I happen to like Foskett, but this treatment was extraordinarily lightweight — less than I would expect from him. It was almost as though he was adhering to Twitter word limits despite the fact that he had at his disposal the limitless page length of the blog.
More to the point, his treatment was a summary of the collective wisdom of the industry, rather than any sort of probing analysis of the validity of the talking points that he was echoing.
Part 1: Disk drives are getting more capacious year over year, while falling in cost. Folks viewed them as “too cheap to manage.”
Part 2: The digital revolution suddenly occured, so there were a lot of files. He calls this unstructured data because he views file systems as the weak sisters of databases, I guess. Users use additional disk capacity like drunken sailors and fill them up with lots of poorly named and unmanaged files because there are no quotas or chargeback systems. Graphics add to the burgeon by increasing file sizes exponentially.
Part 3: Tiering storage enters the fray as a lame alternative to data management, driving up array costs.
Part 4: Operational costs grow with storage capacity.
Part 5: Cloud storage is the solution.
Okay. So, we have a few issues here with Mr. Foskett’s timeline. For example, tiering did not appear in the world because of file proliferation or even after it. It was core to mainframe operations long before so-called open systems were a smile on mommy’s face that daddy didn’t understand.
Second, Foskett completely ignores tape in his tiering paradigm. Big issue that, since using tape is much more cost effective (and trustworthy over time) than using a third party service provider to store all of your extra bits.
Third, he ignores other cost drivers of storage. In addition to the failure to manage data, to implement archives, to engage in periodic data hygeine activity, you also need to factor in OEM practices: huge holdbacks of array capacity, bloated controllers and perTB charges for all that value add software that adds little or no value, obscene mark-ups on commodity disk, hockey stick growth in maintenance contract costs, usual and customary mark-ups by channels who resell OEM gear, no resale value in hardware that has finished its useful life because of non-transferable software licenses (your old NetApp filer is a boat anchor), lack of a common management paradigm across equipment vendors and sometimes in the various lines of products offered by the same OEM, lack of visibility into infrastructure, SAN-everything mentality of the last decade, “converged network” silliness of this decade around FCoE, and extraordinary amounts of dollars spent on misleading marketing by the OEMs and their analyst minions that must be passed on to the very consumers who they are seeking to convince to buy their crap.
Foskett is right about one thing: cloud storage provides a bigger junk drawer under the expensive junk drawer that everyone has already built in their infrastructure. If cloud storage woo is as successful as the purveyors are hoping, consumers will be paying as little attention to the cost and reliability and manageability of this storage tier as they did to all the other tiers.
No wonder the front office hates the back office for its inefficient use of budget and poor return on investment. But then again, in too many companies, management is part of the problem too when they allow EMC, et al, to convince them that their box of hard drives is worthy of a bigger investment than the boxes of disk drives offered by competitor XYZ – when everyone is selling roughly the same equipment.
How about attacking this problem another way? Freeze capacity. Tell everyone they can’t have any more disk, but they can use tape for free. That might just begin the process of culling out the junk drawer.
{ 1 comment… read it below or add one }
Thanks for taking a look at my series which, though published months ago, was referenced last month as part of an email campaign by Nirvanix marketing. Though I suspect many folks will agree with some of it, I hope they won’t disagree with too much of it. And I know it was far from exhaustive – I’m right there with you on the topics of ridiculous margins, add-on software, and foolish technology medicines that cost more than the problem they seek to cure.
And you’re right about tape – properly operated and handled, tape media has always been cheaper than disk and will likely continue to be for a long time to come. But what about all the “woo” that’s added to tape infrastructure? Tape has hugely expensive libraries, operations staff, offsite transportation and storage (that may damage the media), not to mention the crazy-expensive software used to write and manage the data on tape.
Given this, folks have concluded that tape, though certainly cheaper than primary enterprise storage, isn’t quite as cheap as it appears. And what’s to stop even the cheap and well-managed tape environment you envision from becoming another “cheap junk drawer”?
I think both of us would love to see some real progress on storage management, especially categorization and purging of unneeded data. But most sites don’t have that at this point, so storage isn’t getting any cheaper.
You must log in to post a comment.