Claus's Blog

So our buddies at EMC are, as usual, cunning and provocative. Witness storagezilla’s recent post on NetApp’s (still can’t get used to NTAP’s new logo after abandoning what I thought was the coolest logo of any storage company, ever!! (but then I’m not the marketing guy in the room!)) TMS SSD RAMSAN-500 DRAM WAFL EMC PAM SLAM FAM HP CRAM. “Zilla”, ease up on the acronyms or at least append an LOA (list of acronyms) to your posts!! Oh, I forgot, we’re in the storage space…but half your post contained capital letters. I’m not sure if I should be impressed or annoyed…we’ll leave that question for the archives…

But at the end of “Zilla’s” diatribe, he focused a personal question to me, specifically:

“Phrased in the form of a question: Claus, this type of cache due to the fact it is not mirrored is what?”

So, “Phrased in the form of an answer” I would say: “Uhhh, it could be very dangerous?; Uhhh, it could be not a problem?. But it’s basically JBOD (another acronym here so I’m guilty, too).

Here’s the issue, though. The original SSD’s of times past were not only JBOD, they were volatile!! That is, these things were hung off the end of a channel (mainframe, direct attach days, you know, before television was invented) and if you lost power, you lost data!! Just plain old memory. But boy were they fast!! Here’s where the argument converges, though. The SSD’s of old were used for extraordinarily high READ activity (like today’s flash drives should be). That’s where they excelled. Any WRITEs to that memory would be lost in the event of a power failure. But customers KNEW that! So at the end or beginning of a shift, or whatever, they “loaded” the data into their SSD’s and knew if there was a power failure, any changes would be lost. No problem, since the customer’s knew the score.

But this is 2009 (are we really that far past Y2K; seems like just yesterday, I’m getting nostalgic) and we’ve come a long way. Data should be protected and we now have 83 different RAID types, it seems, so NOT protecting data seems a bit silly to me. If a RAID-5 7+1 config increases your costs by a whopping 12.5% on what is already like REALLY expensive media (flash drives), who would ever complain? It’s like Donald Trump complaining that his haircuts (  ) suddenly jumped to $11.25 from the usual $10.00 trim. These days, everything should be mirrored, RAID’d, remote copied, backed up, triplicated and carved in wood. Redundancy is paramount. Redundancy is paramount. Do I need to repeat myself?

So, to answer “zilla’s” question directly, who cares? As long as the customer knows what they’re getting, they’re getting what they paid for. But personally, these days, there is no excuse to not protect data. That’s my personal opinion. And I should say that’s the HDS opinion as well since we do follow the “2-copy model”. Everything is protected, whether you like it or not. No JBOD option, here.

Did I just agree with my mortal EMC enemy? Ouch!!