Meeting Half Way

After today’s Tweet Chat, intended as a promotion for Storage Decisions Chicago 2012, I fear that I may be regarded as a hardcore iconoclast when I present there next week. 

For one thing, I am not sold on clouds for storage (except maybe for longer term archive). 

Nor am I enthralled with Big Data when that term is used to describe a lot of data, rather than analytical tools used to perform well defined investigations of many datasets. 

Nor am I a big fan of FlashSSD, or primary dedupe, or any of the other shiny new things in the industry.

I am prepared to meet some of these technologies half way, as it were.

• FlashSSD is fine for read caching hard disk to give higher IOPS with less power consumption, but I wouldn’t use them as a primary write target in the presence of any serious workload.
• I am willing to embrace cloud backup, provided I can get my data back via tape.
• I am willing to embrace object storage, Hadoop and other methodologies — when they fully mature.

I just can’t see a first adopter advantage in the early versions of these things.  I am especially put off by the marketecture that substitutes for architecture around topics like cloud and big data — tell-tale signs that the tech may not yet be ready for primetime. 

I like my storage technology like I like my wife:  fully matured, well spoken and elegant.  Storage technology shouldn’t be like a box of chocolates — mysterious confections on the inside despite the appealling coating of chocolate on the outside.

See, I don’t think storage needs to be — or should be — a science fair experiment.  It should solve a simple and well-defined problem in a predictable and managable way. 

So, I want uniform management on my rigs, preferably using an open standard like W3C REST. 

I want to virtualize the hardware to eliminate vendor lock-ins and to surface value-add functionality in a way that enables it to be shared across all rigs. 

I still believe that tape has a role in my environment, regardless of the kids who say that tape is old school. 

When you have the same kids downloading the Internet to my disk from every desktop, laptop, tablet, smartphone they can lay their hands on, I call that abuse.  I know their parents used to let them download whatever they wanted when they were growing up, it isn’t a good practice in the workplace.  They need to stop.

I don’t want to hear another person say that they can’t sort out the storage junk drawer because it is above their paygrade.  Print a report of files by user/department and give the lists of the biggest space hogs to their manager with an estimate of how much money it is costing to host all of this seldom-if-ever-accessed data on my gear. 

If they come back with a statement that USER A’s time is more valuable than the cost of the storage, make a note.  The next time senior management wants to cut my budget, I will tell him/her that we could make do with what we have if USER A and his many buddies would clear the clutter from the spindles. 

Truth is, with a bit of data hygeine and an archive (preferably on tape) we could reclaim up to 70% of the capacity we already own.  What would that do to bend the storage cost curve?

While were at it, let’s stop drinking the VMware Kool-Aide — the woo of the V-Partiers. Their stated intention, to add a proprietary storage hypervisor to their proprietary server virtualization microkernel mess, is going to do extraordinary harm to storage.

Think about it:  If you have other hypervisors or other non-virtualized workload using your storage, you are going to end up having to segregate the resources of the “SAN” you just spent ten years and big bucks rolling out, so that the stuff that supports the V-Partiers’ workload is separated from the spindles and rigs that support everything else.  Aggregate, then segregate, then aggravate.

This is a huge distraction from what we should be working on:  holistic management using W3C REST. 

Plus, I can use a server agnostic and storage agnostic virtualization engine like the SANsymphony-V R9 storage hypervisor that DataCore Software will be announcing shortly to improve the operational and financial efficiency of my infrastructure without having to cow-tow to Emperor Vimware at all.  And its a hell of a lot cheaper.

When I say these words, I incur the wrath of VMware fanboys.  When I pose the issues outlined above, their response is, “Well, get rid of all those other hypervisors and non-virtualized workload and just use our stuff.”  I just smile, remembering that this was what Microsoft used to say back in the late 80s, before Samba, when I needed to exchange data between servers running MS and others running Sun OS.  “Well, the solution is to get rid of all that Sun crap.” 

So what if I can’t say that I am on the cutting edge.  From where I’m standing, the more things change, the more they stay the same.

I conclude this rant with a simple observation as a father who is going to be the recipient of some brief homage this weekend from my six kids:  there comes a time when you need to put childish things aside and act like adults.  If it makes me sound iconoclastic, well, there it is.

Hope to see some of you in Chicago at Storage Decisions in the coming week.  And Happy Father’s Day to those blessed with offspring.

Tagged as: deduplication, FlashSSD, Hadoop, management. archiving, RESTful management, Storage Decisions, storage hypervisor, VMware