h a l f b a k e r yYou think: Aha! We go: ha, ha.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
I've been getting fishboned a bit recently and have got to accept that some of my ideas may have been, to use an ego-saving phrase, a little 'controversial'. To save face it's probably about time that I posted something a little more popular. (here's hoping)
Problem: At the moment the only way we
get a faster stream of data from a CD-ROM is to spin the disc faster. The situation can be improved by caching the information but the problem remains one of mechanics - you can only spin the disc so fast without it deforming or your PC attempting vertical lift-off.
Solution: Put independent multiple heads around the CD so that you can:
a) Read items ahead for a cache while still reading the current area of the disc.
b) Rest the heads at different points of the disk reducing seek-time.
c) Serve multiple applications/users simultaneously.
This technology can be applied to just about any disc-based medium and (if baked) certainly hasn't filtered down to cheap IDE Hard Drives. I can hear my hard drive thrashing if it's accessing virtual memory and saving a big file at the same time. Two (or more) heads would solve this. Please let me know if it's baked - I haven't found it yet.
(??) Only link I could find.
http://www.penstars...rives/cddvd_rev.htm Multi-laser CD's. [StarChaser, May 14 2001, last modified Oct 05 2004]
Another link to the Kenwood
http://www.xlr8your...m/G3-ZONE/hival52x/ Multiple beams but not independent. [st3f, May 14 2001, last modified Oct 05 2004]
[link]
|
|
huh. I think rayfo's corrolary applies here, but I can't imagine why it hasn't been done. Would it cost so much more to have multiple lasers or r/w heads and some sequencing firmware? |
|
|
I don't think the heads could be truly independent, but rather, each of the (say) three heads would swing across approximately a third of the disk, without their sweep areas intersecting. It probably would be expensive or unnecessarily difficult to engineer multiple , redundant heads that could each access the entire disk. |
|
|
waitjustaminute... don't most hard drives nowadays have an array of heads along a single arm, thereby requiring much less radial motion? In that sense, I suppose that you could say 'Baked'... |
|
|
Given the plummeting price of memory it might be more practicle to simply have a massive cache. |
|
|
In a couple of years time 640mb of memory may cost the same as 64mb does now. Just stick in 640mb or RAM and have the CD-ROM drive run at 50 speed non stop for a few seconds and load the entiire CD into the cache. |
|
|
I'm sure this could be done now (although not the entire cd, maybe 64mb of it?) Wuold make a good selling point, and cut down the access times. (Many files might not even require the CD to be read.) |
|
|
I think hard drives should have a 64mb internal Cache as well - why the hell not, what's the cost of 64mb of RAM (£15/$20) onto the cost of a hard-drive? Think of the speed improvement! Got be a good selling point. |
|
|
absterge: I keep waiting to hear that it's baked. I had a look at IBM's drives and even the really expensive big server drives have one head per disc surface. Looked at the swing arms too and (without a redesign) figure you could only put two full-sweep arms on one surface. Still, it would be a big improvement. |
|
|
CasaLoco: I just don't buy that. A cache is good but sometimes you're held up by the real speed of the disc. |
|
|
waugs: also applies to DVDs. |
|
|
I think you'll find the heads/swing arm bits etc are the most complex and expensive part of the drive. While it might be a good idea to have more than one - the cost would escalate past practicality. A flash RAM based drive is more likly, or a large non-volitaile DRAM chipset. (both available, but very expensive..) |
|
|
It is actually baked for CD drives. I've seen a 56x drive with apparently 5 laser heads where the drive turns slower because each head is responsible for different sections of the disk. I've seen them in the store, but didn't want to spend the 30$ difference...Now I wish I had. This 56x I have is -noisy-... |
|
|
Starchaser - thanks for the link - I've found another one. It looks like Kenwood have solved the bandwidth problem by using 7 separate laser beams. Trouble is they all come out of the same head, so there's no independent accesses. Pretty close, though. |
|
|
You can achieve something like this yourself; buy 5 CD-ROM drives, get (or burn yourself) 5 identical CDs, and use a software RAID driver configured for "mirroring". The RAID driver will take care of using all the drives at once to maximize read performance and minimize seeks. |
|
|
The difference between this and what you describe is that you're duplicating the spindle, the disc itself and some drive electronics along with the heads, so the net costs more, takes up more space and consumes more power. But, on the bright side, you'll have more reliability in case one of the drives fails, and you can scale the system up or down by adding or removing drives. |
|
|
No problem. The impression I had from the box was that this particular one was multiple laser heads, but I didn't pay that much attention when I saw the '99.99$' sticker... |
|
|
I have a Kenwood 52 True-X. It screams through large files, but seems to have more latency than typical for fragmented searches. Seems to be always a touch out of balance, too.
It was also quite cantankerous to install, requiring me to shuffle my internal zip and HP burner drive letters a few times to get them all happy. |
|
|
As has been pointed out, hard drives use one read/write head per platter surface (typically two per platter, though sometimes the uppermost and bottommost surfaces aren't used), CD-ROMs and some DVD-ROMs do this too. |
|
|
There are harddrives available with 2MB onboard cache which could be increased, bearing in mind that cache memory (unlike RAM) is relatively expensive. And cache memory is only a buffer, so you'd either have to image the media to memory or you'd have to have better algorithms to decide what data to read-ahead and cache to make it efficient. |
|
|
This leads back to the problem with multiple read/write heads: in order to read-ahead, you need some place to cache the data, which means you have to keep checking to make sure you have cache space. You also have to make sure you're caching data that'll be requested. All this stuff adds to the cost of the drive (as [StarChaser] notes). |
|
|
My interpretation is adding say 4 multiple heads to the existing moving assembly of a CD-R, dividing the entire disk into quarters. The performance goal would be reducing seek times and disc-at-once burn rates. It wouldn't help read rates at all unless you used a custom disc format (like RAID per head). |
|
|
That's NOT like a tractor with 4 discs; it's like 4 tractors at once, spread across the field (admittedly welded together, if using the same swingarm). |
|
|
None of the posted objections ring true for me. Caching doesn't improve seek time. A 5-CD RAID can't burn a single CD 5x faster. The "true-X" doesn't improve seek time since all the tracks are next to each other (but it's much more cost effective since there's only one head). |
|
|
Yes, the variable spin rates of CDs make the firmware messy. Also, there would be the added problem of the spiral track: would there be seams at the point where head 4 stops writing and head 5 starts? |
|
|
I've also thought about this idea (thus it's clearly genius :). |
|
|
Of course, the biggest problem is the speed.
As you get further out towards the edge, the cd spins slower. So, unless the laser heads could pivot quickly, you'd have some problems. While laser1 (near spindle) is getting 1 MB/s, laser6 (near edge) is getting 10 MB/s. |
|
|
I think the idea of using this for a hard drive is great! A drive with 5 independant heads on one arm (or multiple arms each with 5 heads) would be great. If the firmware did a logical reordering so the 5 spots that could be read at the same time appeared to be right next to each other to the outside world then 5 heads would roughly make the HD 5 times as fast.
A cache doesn't affect write speed much. RAM prices are doing down but they will probably always be much larger than disk prices and bandwidth needs will continue to grow so I don't think RAM will catch up anytime soon. What happens when you decide you want to edit your home video? |
|
|
I have thought of this as well. The cost added to the hard drive should not be too great. Think of it this way if you want a two drive RAID you are buying two of everything. If a drive has two arms on one drive it can do more than RAID bacause ether head can read any block and yet you only pay for the extra head/arm. It would need a bit more cache because block could be read out of order but must go down the IDE in order. Even with the extra cache it should be cheeper than two full drive if it becomes mainsteam. I also agree there are more problems with CD/DVDs but I don't think they are technically or financially insurmountable. |
|
|
I have also wonders why FAT tables are keep one the hard drive it self. (except for back up) I know the operating system holds a copy but this can be lost in a crash or power lost. Why can't the hard drive itself have flash FAT? Sorry for drifting off the subject but it would be one less place a head would need to go. |
|
|
I like this idea and can even imagine it from an engineering standpoint. I taken apart several drives. To economize space arm assemblies are put in one corner of the drive. Put 'em in the other corners. One down side might be the additional where on the disk surface.Conversely another benefit would be that heads would be redundant in case of failure. I've also seen some DJ's with some goofy turntable with two arms that work on the same idea. |
|
| |