August 20, 2007 at 5:14 pm
Actually most people I know in IT could use a maid, but I'm talking about MAID, massive array of idle disks. Seems like just when I thought all the X-Array-X-Disks acronyms were used, here comes another one.
It's a technology that could save power and handle large data loads. It uses large amounts of idle, non-spinning disks with a smaller number of spinning disks that act as a cache. I guess it's not new, though I've not seen it in the news or heard of it before.
As power costs grow, however, it's apparently getting some interest. After all, there's a good chance that a large amount of our data is rarely or even never accessed, but it might be and it needs to be kept in a form that can be accessed.
Forgetting about the possible problems of restarting those disks, this sounds very much like the old Hierarchial Storage Management (HSM) systems I worked on early in my career. In those systems we moved data from disks to optical media. We used jukeboxes (does anyone under 25 know what those are?) that held large WORM or rewriteable optical media. I guess large is a relative term since the storage capabilities of today's flash drives exceeded what our 8 inch opticals held.
The idea is the same, and I'm not sure which is a better one. Disks could continue to be expanded, in a SAN like fashion, though I suppose some sort of tape or optical library could do something similar. Either way it's probably an investment in a vendor as much as equipment as the conversion costs could be huge.
It's also a big job in determining what content would be migrated. While it's mostly automated, I'm sure that someone needs to oversee the process and correct any mistakes that the algorithm might make. It used to be rare that the system made mistakes in moving or retrieving data, but rare isn't never.
The bigger problem is in managing the actual storage devices. It's easy to let them fill up, somehow get unbalanced, etc. Especially as when the system goes in, the devices will be mostly empty. Which means you won't have anything to pay attention to for awhile. It would almost be better if you could buy a small system to start managing a high percentage of your data and then grow it on a regular basis, like quarterly, so it's not out of use, out of mind.
I just think that as long as I'm not the maid for a MAID system, it sounds like a good technology to implement.
Steve Jones
August 20, 2007 at 8:29 pm
I think I'll just stick with my MAID... just to make things clear .
August 20, 2007 at 9:03 pm
I thought you are talking about me...... Database developer during the day, the maid of the my family when I am not working at my office - that is the job which has no pay, no one appreciates, 7 days a week, 24 hours a day, duties include cooking, cleaning, laundry, .....
Hey, there is one thing in common between my database developer job and the maid - no matter what I do and how hard I work, no one appreciates and expecting me doing more and more.
August 21, 2007 at 4:56 am
No offense Loner, but you really need to go out and Watch the movie "The Secret".
Just trust me on this one .
August 21, 2007 at 6:33 am
So they're using SATA drives to create these arrays. Right? Here's my beef with this technology and I'm sure they have some way of dealing with it, but what happens when the disk won't start back up?
I'm not sure how many of you remember a while back like 10-15 years I guess now, but Dell bought a whole bunch of Western Digital HDs. Apparently the wrong type of grease was used in the ball bearings or something, so that when the drives were powered down and stopped spinning they froze and wouldn't start back up...
Like your car the most things that can go wrong typically occur at startup. I understand the whole cost savings bit on AC and power, but it seems to me you'd be inviting a large amount of potential risk by using this type of platform. When you do a risk reward analysis, unless the vendor can in some way guarantee the disks and the data stored on them with some sort of amazing SLA I'd be hesitant to consider this for anything beyond a simple file share...
August 21, 2007 at 6:49 am
Luke,
That was my first thought, but I'm assuming they have some sort of RAID technology as well built in. The MAID software also periodically (I think I saw weekly from one vendor), exercises the disks and reports on issues, moving data to a new disk.
I'm skeptical, but there are some smart guys and gals that come up with how this stuff works and I hope they've considered disk failures.
August 21, 2007 at 6:57 am
Sure, and I get that, it's just me perhaps being a bit over paranoid, but it comes down to what you said at the end of your editorial, I wouldn't want to be the maid taking care of a MAID. Even if there is redundant disks and disk exercises and reporting and such, I would imagine there would still be a much higher disk failure rate than in a normal SAN. Which means more work for me, or my server guys or whomever, and it also means going back to the vendor and getting new disks (yeah, yeah have some extra on hand, but what if...). I guess that was more what I was getting at.
August 21, 2007 at 7:58 am
Those are definitely good points to consider. Especially with large SANS dropping disks on a weekly basis in many places.
I'm definitely not sure if I'd want to take care of this beast. Maybe in 10 years, sip my coffee, do the crossword and change a disk every 15 minutes
August 21, 2007 at 8:46 am
Hmmm. Never heard of it. I, too, don't like the potential risk of drive failure from these things being powered down for so long, even if they are occasionally exercised. The article mentions dual fiber channel connectors, which means those are going to be expensive drives and controllers, therefore the whole thing is going to cost a lot.
I can see the heat savings. We're in the process of reevaluating how many BTUs our current equipment produces in preparation for a new building that we'll be moving in to in three years, but at that point our S/390 will be retired and that will recover A LOT of cooling capacity! Still, especially these days, we've got to keep track of BTU production and KW/hour consumption.
I can't see this as a solution for us as it doesn't solve any problems that we have. For other industries, sure. The article mentioned a petabyte for a medical research outfit. It could be a good fit for them, but I don't think it's a good fit for everybody.
-----
[font="Arial"]Knowledge is of two kinds. We know a subject ourselves or we know where we can find information upon it. --Samuel Johnson[/font]
August 21, 2007 at 11:01 am
Yesterday I was researching for a server power supply replacement and came across a blog about some guy buying one lot (10 or more) of 10 gig drives and was happy about it.
I thought he was mad.
After this article, maybe he was just MAID!
groan
I don't see how this is energy efficient or saves energy unless they run off of solar cells powered by the overhead lights. The general theory of consistent starting and stopping of motors is more inefficient and cost more long term may not be correct but certainly leads to premature failure of the motors.
Re: Mythbusters show about leaving lights on versus turn them on and off as needed. The cost savings of leaving the lights on was insignificant to the start up cost. But these are motors and not light bulb filaments.
Viewing 10 posts - 1 through 9 (of 9 total)
You must be logged in to reply to this topic. Login to reply