Slashdot Mirror
Miniature 5400 and 7200 RPM HDDs Reviewed
PReDiToR writes "At Tom's Hardware I found this favourable review of some remarkable Hard Drives. The article points out that with 40GB units suitable for server or desktop use, life with 2.5" drives could be just around the corner. Heat noise and power consumption are all apparently within acceptable tolerances."
What's impressive to me is that Tom's HG is still this fast , even with all of us checking out the drives. Tom must be the hardware king...
/. is still quoting Tom's HG here. The last several articles I have read have caused me to lose all respect for them. All fluff, no facts, lots of generalizations, and no real useful information. This has been discussed here previously, and many are like me, and no longer bothering reading Tom's reviews.
What impresses me is that
They used to be the king, but the king has no clothes. Sorry about the OT, but it's relevent enough if it keeps anyone else from wasting their time.
Tequila: It's not just for breakfast anymore!
...I just got a WD Raptor 10,000rpm SATA drive witha 5 year warranty (under 100UKP) for my desktop. Try and keep up, people!
I believe that's 5400 rpm...
Hier staat een stukje tekst.
On the other hand, smaller drives are likely to be more durable. Smaller platters means shorter arms, which means less "flapping" if you apply a shock to the drive.
As far as the "abuse" is concerned, I think head crashes are a greater danger than bearings dying.
Tarsnap: Online backups for the truly paranoid
You're limited to 1.8" drives, which basically means Toshiba, topping out at 30GB with the MK3004GAH, a 4200rpm unit with slow access.
Most (if not all) TravelStars use FDB technology now--basically they use oil in the place of little steel balls. It reduces power consumption and makes the drives much quieter--and more stable at speeds.
Q: "Why do sound techs say 'check 1, 2'?"
A: "Cause if they could count any higher they'd be lighting techs."
raid arrays with usb based drives...
Didn't someone do this already?
AFAIK the transfer speed of 10,000 and 15,000 SCSI drives isn't so far ahead of the 7,200 offerings, the improvement is in the latency. I have a SCSI system here at home and the combination of low-latency and tagged command-queing (buffering of future disk requests) lets applications start up lightning fast. Loading video and MP3 files is actually faster off the ATA drive though, as it has a higher raw transfer speed.
What would be really useful for workstations in this day and age are 16MB buffer 2.5" 5400 RPM disks running SATA with TCQ. An intelligent pre-buffering system to keep the 16MB buffer full of anticipated data would help the drives perform as well as 7200 RPM drives, and the lower heat generation and power requirements would boost corporate adoption. I'll bet that after a short while these drives could even be cheaper to produce, as there would be about 1/6th the raw materials needed; the only reason they cost more now is the lower volume.
"Sometimes, I think Trent just needs a cup of hot chocolate and a blankie." -Tori Amos on Nine Inch Nails
the only reason they perform so badly is because there's no decent buffering in ATA drives. With Tagged Command Queing and big prefetch and write buffers they could perform well at all but transferring very large data files (not a common case on most desktops, BTW).
Also, why don't modern operating systems aggressively prefetch file system metadata and keep it in cache? It seems to me that with most systems having more than adequate RAM it would make sense to keep the entire directory listing and metadata cached. I think Reiser4 has a method to determine how 'hot' a directory is based on how often it changes, there should be a way to tell the kernel you want to aggressively prefetch indexes for all objects on '/dev/hd?' with 'heat=>2'.
"Sometimes, I think Trent just needs a cup of hot chocolate and a blankie." -Tori Amos on Nine Inch Nails
For the price you pay, 2.5" drives are horribly inefficient, and nowhere near as fast as 3.5" models.
Talking about the models reviewed, yes, but that's soon to change, very soon. Seagate's Small Form Factor drives will be around next year. At 10,000 rpm and with a U320 SCSI, Fibre Channel, or Serial-Attached SCSI interface, they're as fast or faster than most of the 3.5" drives out there. The platters in the 3.5" enterprise drives are as small as the ones in the 2.5" anyway, and you'll (almost) be able to fit 4 2.5" drives in the space of one 3.5".
Sorry, can't quite follow you about the 10V thing. Yes, first generation flash chips had a Vpp of 12V, but that was about 15 years ago. The later chips all have integrated charge pumps to generate whatever voltages they need internally, so it's completely transparent to the circuit designer.
It's not about remapping the sectors, it's about distributing the wear so that all sectors get worn out equally. Once you get an error, the flash chip is probably already breathing its last. The point is that it takes so long for all sectors to wear out that it doesn't really matter.
The card I used had my calendar, contacts and database files on it, which got modified (i.e. written) a lot during a business day.
Flash isn't really suited for replacing HDDs in PCs, as I said, but this thread was about USB keys vs. hard disks for carrying around storage, and my point was just that flash cards or sticks or whatever are better suited for that than hard disks, and of course that the flash memories wearing out quickly is a bit of an urban legend.
Did you know you can fertilize your lawn with used motor oil?
Ok, here's a positive reply.
Ive started on this very thing myself, using firewire however.
Unfortunatly I started getting all sorts of ideas and tacked them on, and now 'cheapie' is out of the question.
I built a system with 4 firewire buses just for disks.
I also chose not to power the disks from the firewire bus (explained bellow why)
The master plan is to have 6 firewire buses (two groups of three, and they only had dual bus cards, thus why i have 4 ports now.. One is not used yet, and will be used with the 3rd dual fw card i do not yet have)
These three buses connect to firewire hubs.
Then, you connect three disks per hub.
Now is the confusing part.
On bus A, you have 3 disks. These are disk A from 3 different raid-5 groups.
Bus B has 3 disks which are disk B of the same raid-5 groups. and so on for C.
Then you setup a raid 5 group out of just those 3 disks. Which in this case gives me 3 groups, and with the 3rd firewire card (bus 5 and 6) this concept will double.
Then you use LVM to link all of the raid-5 groups together into one big volume.
Reasonings:
If any one firewire bus failed (or was unplugged) only ONE disk from each raid-5 group is offline. Raid-5 can survive this.
I dont use power over firewire because a) the PCs supply can not handle all of that, and b) I now have 3 power supplys, each one chained to the disks in the same order as the firewire buses. This way if a power supply failed, it only takes down one disk per raid-5 group, and again raid can survive that.
Firewire is multi-host (IE you can have more than one host controller on the same bus) so with two computers on the bus, doing heartbeat monitoring over a serial link between them, if a computer failed, the other can pick up the disks on the bus and continue file serving.
Using LVM to link the raid 5 groups together means after i start getting disk failures 4-5 years down the road (well, hopefully that long) and it starts to get hard/expensive to find disks of the size i am currently using, I can move the data off the raid-5 group to unused space, and decomition that group. Then it can be pulled off the bus, and replaced with current newer disks which ideally will be much higher capasity, without replacing ALL the disks in my array (as would be the case with a single raid-5 array of all the disks)
Then you recycle the failing disk, and have two disks spare to use for other machines (IE a mirror to boot a new machine off of, spare single disks, etc)
Some links you may find interesting:
- http://www.fwdepot.com/
Best source of firewire controllers, bridgeboards (firewire -> IDE, firewire -> scsi, usb->ide, enclosures, clamshells you mentioned, etc etc)
- http://evms.sourceforge.net/
EVMS = Enterprise Volume Management System. Linux software that lets you manage raid, lvm, clustering, etc all from a server setup. Comes with cli, curses, and X11 interfaces. Not quite 100% there yet, and still has a couple problems for enterprise use, but almost all of them are related to the 2.4 kernel and promised to be fixed when 2.6 is out/stable (and in the past month very well could have been, i havent been keeping up)
(Please please dont slashdot my poor little file server here!
If anyone would like to mirror, its ok with me. This is a p2 200 and will die if more than a few peope hit it at once)
- http://photo.brokensphere.net/index.cgi?mode=view& album=home/fileserver
(Watch for slashdot injected spaces!)
This is pictures of some of the parts at the start of my project.
Havent added new picts yet, nor had much time to work on it.
These picts show the disks all on one bus and interlinked, which is not good for speed, but I was just testing the EVMS software at the time.
In the end, I plan to make my own case, which may be a sheet of half-inch think wood screwed into a wall, with disks hanging on it as so their tops face out, and a plexigl