<strong>P.S. And for those who smoke in the presence of their computer or at leat their exposed hard disk (cover off etc). As the disk heads only sit about 15 micron from the surface of the disk, please don't smoke, a smoke molecule is about 80 micron and yes it can cause a drive crash if you blew smoke onto the drive.</strong><hr></blockquote>
Um, every last drive I've ever looked at had a stupidly hard-to-remove dust-sealed cover replete with stickers promising vast amounts of nasty stuff being visited upon you and yours should you ever ask for warrantied repairs once they're less than intact, not to mention the funny condom-like rubber shockproof cases on all these Seagates I have kicking around here. Oh, and most drives I've had to bits wouldn't actually work if you took the top cover off, since the head mechanism pivot is secured to it...
If smoke particles are getting anywhere near the platter, you've been doing something distinctly unauthorised...
<strong>I just had a thought. Instead of spinning drives faster, why not add a second or third read/write head to each platter? iDunno how any of it would work, but a second head should make it possible for the disc to only need to spin half as much in search of a file, no? Ah well, there are probably other issues besides cost.</strong><hr></blockquote>
At first I thought this idea was pretty useless. Current drives utilize multiple platters stacked on top of each other with a single pivoting arm of heads suffled betweeen the platters, moving in unison. These things are fairly delicate and move back n' fourth fast fast fast!
But here's the idea, stick another set of these on the opposite edge of the platter accross the spindle. Since each arm slides in from the side, perhaps they may be aligned so that they don't collide. This would reduce seek time by half on average since the platter would only need to spin half way around to hit the data versus all the way around in the worst case scenario.
(Ok, well the worst case scenario might be more than 1/2 around, but its obviously closer than all the way round! )
Anyhow, if you can't visualize this I'll toss together a quicky graphic. It would kind look like a record player with a second arm mounted across from the first, as if the arm was rotated around to the other side of the record.
At first I thought this idea was pretty useless. Current drives utilize multiple platters stacked on top of each other with a single pivoting arm of heads suffled betweeen the platters, moving in unison. These things are fairly delicate and move back n' fourth fast fast fast!
But here's the idea, stick another set of these on the opposite edge of the platter accross the spindle. Since each arm slides in from the side, perhaps they may be aligned so that they don't collide. This would reduce seek time by half on average since the platter would only need to spin half way around to hit the data versus all the way around in the worst case scenario.
(Ok, well the worst case scenario might be more than 1/2 around, but its obviously closer than all the way round! )
Anyhow, if you can't visualize this I'll toss together a quicky graphic. It would kind look like a record player with a second arm mounted across from the first, as if the arm was rotated around to the other side of the record.</strong><hr></blockquote>
isn't this essentially what RAID does, except built into one disk. How much faster is RAID (?25%?). Sounds like a good idea to me. Wonder why they haven't done it (cost?)
[quote]Makes me laugh when I see "the biggest economy in the world" tagged onto the US, not since January 2002 it's not! <hr></blockquote>
Got a link?
It was my impression that Euro-land was not only poorer than the US, but that due to its lower productivity growth, was going to be progressively (relatively) poorer for the forseeable future. Some more on both GDP growth rates and Germany in particular:
RAID is great for data integrity and/or increased bandwidth, but it definitely doesn't do anything for latency. If anything, some flavors of RAID (RAID 5 in particular) can add latency.
The major reason for moving to 10k and 15k RPM drives is that seek times (a measure of latency) are inversely proportional to the rate of spin- so the faster the spin, the lower the seek time.
For the SCSI vs. IDE argument, remember that SCSI can also support command queueing, so the host controller can reorder commands on-the-fly for better performance under a heavy load.
Of course, once SerialATA comes out, we'll have to re-evaluate the whole situation, but we've got a year or so there.
Comments
<strong>P.S. And for those who smoke in the presence of their computer or at leat their exposed hard disk (cover off etc). As the disk heads only sit about 15 micron from the surface of the disk, please don't smoke, a smoke molecule is about 80 micron and yes it can cause a drive crash if you blew smoke onto the drive.</strong><hr></blockquote>
Um, every last drive I've ever looked at had a stupidly hard-to-remove dust-sealed cover replete with stickers promising vast amounts of nasty stuff being visited upon you and yours should you ever ask for warrantied repairs once they're less than intact, not to mention the funny condom-like rubber shockproof cases on all these Seagates I have kicking around here. Oh, and most drives I've had to bits wouldn't actually work if you took the top cover off, since the head mechanism pivot is secured to it...
If smoke particles are getting anywhere near the platter, you've been doing something distinctly unauthorised...
<strong>I just had a thought. Instead of spinning drives faster, why not add a second or third read/write head to each platter? iDunno how any of it would work, but a second head should make it possible for the disc to only need to spin half as much in search of a file, no? Ah well, there are probably other issues besides cost.</strong><hr></blockquote>
At first I thought this idea was pretty useless. Current drives utilize multiple platters stacked on top of each other with a single pivoting arm of heads suffled betweeen the platters, moving in unison. These things are fairly delicate and move back n' fourth fast fast fast!
But here's the idea, stick another set of these on the opposite edge of the platter accross the spindle. Since each arm slides in from the side, perhaps they may be aligned so that they don't collide. This would reduce seek time by half on average since the platter would only need to spin half way around to hit the data versus all the way around in the worst case scenario.
(Ok, well the worst case scenario might be more than 1/2 around, but its obviously closer than all the way round!
Anyhow, if you can't visualize this I'll toss together a quicky graphic. It would kind look like a record player with a second arm mounted across from the first, as if the arm was rotated around to the other side of the record.
<strong>
At first I thought this idea was pretty useless. Current drives utilize multiple platters stacked on top of each other with a single pivoting arm of heads suffled betweeen the platters, moving in unison. These things are fairly delicate and move back n' fourth fast fast fast!
But here's the idea, stick another set of these on the opposite edge of the platter accross the spindle. Since each arm slides in from the side, perhaps they may be aligned so that they don't collide. This would reduce seek time by half on average since the platter would only need to spin half way around to hit the data versus all the way around in the worst case scenario.
(Ok, well the worst case scenario might be more than 1/2 around, but its obviously closer than all the way round!
Anyhow, if you can't visualize this I'll toss together a quicky graphic. It would kind look like a record player with a second arm mounted across from the first, as if the arm was rotated around to the other side of the record.</strong><hr></blockquote>
isn't this essentially what RAID does, except built into one disk. How much faster is RAID (?25%?). Sounds like a good idea to me. Wonder why they haven't done it (cost?)
Honest question here-
[quote]Makes me laugh when I see "the biggest economy in the world" tagged onto the US, not since January 2002 it's not! <hr></blockquote>
Got a link?
It was my impression that Euro-land was not only poorer than the US, but that due to its lower productivity growth, was going to be progressively (relatively) poorer for the forseeable future. Some more on both GDP growth rates and Germany in particular:
<a href="http://www.economist.com/displaystory.cfm?story_id=1464893"; target="_blank">http://www.economist.com/displaystory.cfm?story_id=1464893</a>;
<a href="http://www.economist.com/agenda/displayStory.cfm?story_id=1477338"; target="_blank">http://www.economist.com/agenda/displayStory.cfm?story_id=1477338</a>;
In all honesty, I have to admit to being puzzled that an Englishman would be interested in standing up for Germans:
<a href="http://news.bbc.co.uk/2/hi/uk_news/2558083.stm"; target="_blank">http://news.bbc.co.uk/2/hi/uk_news/2558083.stm</a>;
On topic:
RAID is great for data integrity and/or increased bandwidth, but it definitely doesn't do anything for latency. If anything, some flavors of RAID (RAID 5 in particular) can add latency.
The major reason for moving to 10k and 15k RPM drives is that seek times (a measure of latency) are inversely proportional to the rate of spin- so the faster the spin, the lower the seek time.
For the SCSI vs. IDE argument, remember that SCSI can also support command queueing, so the host controller can reorder commands on-the-fly for better performance under a heavy load.
Of course, once SerialATA comes out, we'll have to re-evaluate the whole situation, but we've got a year or so there.
Hope this helped,
-HOS