You *can* have it all.
Nigel over a Ruptured Monkey shows us his Baboon ass and then proceeds to fling dung in our general direction. Well it could be worse, he could be drawing images of his hero Hu Yoshida again. When I first saw that image I had to ask myself why Santa would be going from datacenter to datacenter trying to lobotomize intelligent storage arrays when he should be out getting me a Nintendo Wii. ;-p
Enough joking with Nigel now down to work, there's a lot of other vendors mentioned in that post and from experience their answer to the double disk data loss question is usually the same. "Use RAID 6".
EMC on the other hand reached the same fork in the road all the other guys did but took a different path. The EMC answer as I see it is: "We've enhanced array based protection against data loss due to double disk failures and unnecessary rebuilds via proactive hot sparing, rebuild logging, and other array based error detection and correction techniques, regardless of the RAID type."
Not as snappy as Use RAID 6 I'll admit, but it gets the point across.
Putting on my Mr Obvious hat, RAID 6 does nothing to protect your existing non-RAID 6 LUNs from data loss due to double disk failures, I know you read these posts for such stunning insights but someone has to say it, so why are EMC the only vendor mentioned spending a lot of time and effort working to mitigate the chance of Double Disk failure related data loss across *all* the LUNs in the array regardless of RAID type, instead of just telling customers "Use RAID 6, you're a sucker if you don't"?
RAID 6 isn't hard, it's not rocket science, it's not magic and as it has been pointed out by Nigel it's been done. I don't think it's been done as well as it could have been, but done in some fashion none the less. RAID 6 just doesn't address the double disk failure question in totality. When someone gives you a "Just use RAID 6" answer it means that they either can't or won't do the hard work, do the rocket science, make the magic, required to slay the double disk failure dragon. They're not willing to make all LUNs more resilient, just a specific type of LUN.
This isn't JBOD, the array should work to mitigate the risk of a double disk failure regardless of the RAID type of the LUN, and that's what EMC has brought to market while competitors have been doing nothing but banging the same RAID 6 drum.
Comments
Cheers for your pennies worth though - interesting and educational as always. How long before EMC do start offering RAID 6? hmmmmm
Great blog keep it coming!
As for your question about RAID 6, I'll never answer any questions like that until I know for a fact that EMC has already answered it publicly. To my knowledge EMC continues to evaluate RAID 6 as well as many other data protection technologies..
I know, I'm no fun. ;)
We all know that EMC² will come out with RAID 6 just to keep up with the Jones's. They used to have their own implementation of a RAID 5 knock off called RAID S, now they have RAID 5 and no RAID S. They used to tell you "you don't need mirrored cache", now they have mirrored cache. They used to tell you "you don't need RAID 6" and pretty soon they'll release it. It's always the same story with EMC², tell us we don't need it or come out with something proprietary and then come out with what the rest of the industry has.
Thanks, Snig
As for vendors improving their code not all levels of effort are equal, the difference being in this case that EMC specifically devoted significant resources to those areas instead of keeping up with the Jones's as you put it and dropping in a commodity feature and calling it gravy.
Cheers and Merry Christmas
And just because EMC² has to spend a bunch more money than everyone else to fix the code they wrote/broke in the first place doesn't mean they are improving the end users life due to their investment. They are trying to fix issues that they themselves created and call it R&D. Write good code from the beginning and you won't have to roll out new fixes once or twice a month.
I'm not saying any such thing Snig as I haven't been involved in any conversations you've had, nor am I going to take responsibility for what people in a 31K employee organization may or may not have said at some stage or another. If you're looking for someone to do that you've come to the wrong place. I've told you what the company has said, if people aren't they should be. If they think RAID 6 is wrong for their customer then of course they should say so.
Your point on software and fixes doesn't tally when the reality of software development is that *all* code is broken. Maybe if Donald Knuth was writing every line of code in every product there would never be a need for fixes for anything, but he isn't. I always find it startling when someone claims that if xyz wrote “good” code in the first place they wouldn’t have to fix it, it shows me the person has never had to write a major piece of software in their life and has zero understanding of what’s involved. Showing my UNIX admin roots look at BSD or any GPL’d product some of the smartest developers in computer software built those foundations from principles developed in the 70’s, if not earlier, and to this day their code is still modified, re-written, fixed, and improved upon in a regular basis.
Production code is organic; it needs continuous attention in order to keep it relevant and healthy.
Since I'm a consumer of EMC's storage products, just like you Nigel, I'm probably not the best person to walk people through the Microcode development cycle. That and the fact that I don't want to say anything I probably shouldn't. I'm a guy with a blog, not EMC's voice after all. ;)
It's a bit of a shame that some EMC guys have chosen to not blog at moment as I can think of a few people who'd be great to hear from on this. It's also true that they'd probably blog in assembly language, but yeah. :)
Blog in assmebler - hah assmebler is for WIMPS!!! I prefer pure binary myself ;-)
Oh and of course I'd include parity at the end of each line of my blog just in case some got lost/corrupted. In fact I'd probably include double parity as my blog is sooo important that it would be a disaster to the posterity of the storage world if any of it were lost. Although on the downside that would confuse the EMC world as you've not got round to double parity yet (sorry could't resist that one!!).
Seriously though I agree that it would be good if more EMC people were blogging and talking about stuff like this - they're probably still "evaluating" blogging though ;-)
Cheers
EMC is well past evaluating blogging, there are more blogs knocking around the place but they're internal. The difficulty has been in finding people who were interested in blogging about the information industry, as either as a face of the company or just for themselves. Mark Lewis, Chuck Hollis and the RSA Bloggers are faces of the company while Josh Maher and I blog for ourselves.
A lot of people are either too busy, too shy, or just not interested in keeping a blog. I'd guess, and it is a guess, that there will be more EMC face of the company bloggers pop up though.