We talked to four Reg-reading storage gurus on where and when to use 3.5-inch hard disk drives and when to use 2.5-inch ones. They each admit a role for the smaller form factor drives – with one view being that they could take over completely as their capacities increase. But not one of our quartet thinks that 2.5-inch drive use …
Easy selection method
An easy way at the moment to decide for an SMB requirement where a single SAN is going to provide all the shared storage is to ask "will you need bulk storage/archive/backup to disk?".
If the answer is yes, the capacity uplift you get from 3.5" SATA/NL-SAS disks capacities being up to 2TB compared to ~500GB on 2.5" makes the decision relatively easy.
Double the bays with 2.5" chasiss x 1/4 of the storage density = half the capacity of a disk 3.5" solution. Spindle count tends not to come into it for bulk storage, but you do use an extra bay to go RAID6 for high capacity disks, but you still come out ahead 9 times out of 10.
So, if you need a mix of fast and bulk storage, the SAS disks should be 3.5" - if you don't need bulk you can go 2.5"....
RE: Double the bays with 2.5" chasiss x 1/4 of the storage density
If you've been paying attention its more like triple the drive bays, and 1/2 the storage density.
2.5" Performance Myth?
I'm not saying that 2.5" HDDs will not take over. I am, however, of the mindset that it will be because of power and density reasons and not necessarily performance. I know I am an oddball in that the I/O I deal with is effectively completely random and does not play well with hierarchical storage schemes (or SANs for that matter) but the last time we went looking at IOP specs the absolute best HDD performance we could find 3.5" was still significantly better (to the tune of 20+% IOPS/spindle) than its 2.5" counterpart. The strange part was that the storage consultant from the vendor we were dealing with argued with us about it repeating some of the same 2.5" performance is better lines I see here. We had to show him his own company's IOPS spec sheets (which are a serious pain to find BTW) at which point he 360'd on us and started agreeing with us.
Granted, there is a lot of confusion here and what I saw (4-5 months ago) might not still be valid, but the industry needs a lot of improvement in how they describe requirements on the system/software side and performance on the hardware side. This is 2011 and we might as well be using tea leaves, newts and rolled bones to guess at how particular systems will behave with particular storage architectures.
Small Disks = Better Perf
In my opinion, the smaller the disk size, the faster the data gets read per rotation, less power consumed and smaller seek times.
Not quite so simple
To get to, say, 15k RPM the outer part of a 3.5" drive travels much faster than a 2.5" HDD. Evan Unrue explains this quite nicely - sequential reads on the outer parts of 3.5" drives theoretically should be better than 2.5" drives but for random IO the arm in a 2.5" HDD doesn't have as far to move so theoretically it should be better in that scenario.
BUT... at the end of the day, for now, based on what I've seen... whatever inherent performance differences exist between the two I have yet to see hard evidence that: 1.) the advantages/disadvantages are sufficiently reliable and consistent across manufacturers or that, 2.) given what consistency can be surmised from the mfr's specs you can find, it's significantly different than (at best) a wash from a performance standpoint between the two. You can, quite readily IME, go out to a particular mfr and see examples of comparable (RPM, interface, storage) 2.5" and 3.5" drives where the 2.5" wins... and plenty where the 3.5" still wins.
The current situation could very well just be the result of 3.5" being slightly more evolved and tweaked, but still there are plenty of other really good reasons to like 2.5" and to choose it over 3.5". Just don't kid yourself into thinking that the form factor in either case buys you a significant performance difference. When it comes to HDDs it's still almost all about interface, RPMs and number of spindles.
more rotational delays waiting for multiple sectors to appear under the heads....
Not cut and dried by any manner.
I've started recommending 2.5" drives for backup purposes (Time Machine, retrospect, etc) to friends. Often they have plenty of capacity for their needs and the benefit of a pocketable enclosure (tidier, might actually hide it away or rotate them) and no need for a power supply (tidier, cheaper to run) outweight the cost/capacity disadvantage.
Hot Data: 10K 2.5" Drives
Cold Data: 5.4K 3.5" Drives
There's a lot more cold data than hot on my network. Get the right controller and you can spin down your 3.5" disks when not in use. (I do like MAIDs.) Newer controllers (LSI has sever very nice ones doing 6GB SAS now) can recognise and deal with SSD/Hot 2.5"/MAID 3.5" disks separately and in different appropriate fashions for dirt cheap.
It is understandable for you to be using 3.5" drives right now if you are carrying over legacy equipment (especially if you are towards the end of your refresh cycle,) but by the next refresh, there really will be no excuse.
...but as an enterprise guy who has dealt pretty much exclusively with SAS and SAN, where could I do more reading on MAIDs. I saw you reference that in another comment recently and my Google search really didn't find much on it other than a three-or-so-paragraph Wiki article. I really like the concept and would like to read more on it. What software/hardware do you use for MAIDs? Can a normal RAID setup be used as a MAID?
Hate to sound like a NOOB here but I've been in the business for 15 years and this is the first time I've run across it.
MAIDs are for all intents and purposes standard RAIDs wherein the disks are "spun down" when not in use. That is 100% in the RAID card. The Intel RS2BL080 is my current favourite card. It uses an LSI 2108 chip. It seems to spin my disks down when idle just fine. (I believe you need MegaRAID 3.6.)
DELL PERC H700 and H800 card can also be configured for Spin Down.
Most vendors don't call it MAID. They just call it "spin down." Usually tout it as a power saving feature. I should point out that a single modern RAID card can be married to SAS expanders to provide truly a "Massive Array of Idle Disks." :)
Thanks for the info :)
IOPS or IOPS/TB?
Regarding IOPS and different drive types, looking at disks individually doesn't give the best picture. For example if you take a 3.5" drive and a 2.5" drive, both with 600GB capacity but the 3.5" spinning at 15Krpm and the 2.5" at 10Krpm, you can get approximately 25% more IOPS from the 3.5" drive.
When you move to a proper storage array things change because you can pack so many more 2.5" drives in to the same space. This increases the absolute number of IOPS but not the IOPS/TB.
So when considering any (non-streaming) storage system you need to have three technical metrics in your head: TB, IOPS and IOPS/TB. Once you have an idea of what these numbers need to be you can start to make a decision between different hard drive form factors, capacities and rotational speeds. Other metrics such as cost, maintenance over useful lifetime (==cost), space taken up (== cost) and power required (==cost) probably also matter to you, and again depending on your requirement one drive type will be cheaper than the others.
You can also use this to decide if SSD is worth it for you as well.
Chris Evans' comment "data loss is only an issue if two drives fail in the same RAID group" is a bit worrying coming from a professional storage consultant. RAID 5, 10, fair enough, but RAID 6 and 60 will keep on rocking with two failures, or four in the case of 60
Another benefit of 2.5 HDD is they have much higher MTBF, about a third higher than 3.5 drives. Mind you, thats if you place any faith in them, I'm more of a subscriber to the "MTBF=Meaningless totally bullsh*t figure" theory, but to be fair the lesser head movement should definately translate to longer life
- NASA boffin: RIDDLE of odd BULGE FOUND on MOON is SOLVED
- SOULLESS machine-intelligence ROBOT cars to hit Blighty in 2015
- BuzzGasm! Thirteen Astonishing True Facts You Never Knew About SCREWS
- Worstall on Wednesday YES, iPhones ARE getting slower with each new release of iOS
- Tor attack nodes RIPPED MASKS off users for 6 MONTHS