scottc@bnlux0.bnl.gov (david scott coburn) (09/20/90)
There seems to be a constant barrage of numbers in the trade press and advertising concerning data transfer rates for various types of hard disk interfaces (MFM, RLL, ARLL, IDE, ESDI, SCSI, etc). Can anyone shed some light on this subject? It seems to me that the limiting rate (assuming a fast enough processor) would be the data rate coming off of the head-disk assembly. For example, MFM seems to be limited to 5 Mbits/sec, versus RLL which is 7.5 Mbits/sec. This makes sense, since RLL has 1.5 times the linear bit density on the disk. I don't know how ESDI records on the disk. On the other hand, I have seen some SCSI interfaces (the controllers, not the disks) that claim 16 or 32 Mbits/sec data rates. Wouldn't the SCSI transfer rate ultimately be limited by the recording method? In other words, why should the data transfer rates for a RLL disk with a host-resident RLL controller differ from a RLL disk with an embedded SCSI controller? Isn't this just a case of having the embedded controller AND the host processor wait for the data to come off the disk, rather than just the host processor? I realize that there are a lot of factors that go into determining data _through-put_ rates in a computer system (the io bus transfer rates, on- board cache (on both controller cards and the embedded 'smart' drives), efficiency of the software drivers, the mix of requests to the drive system, etc). If someone is aware of an article or such concerning comparative performance of the many different types of drive interfaces I would be interested in hearing from you. Is there also a reasonably technical source of information on the inner workings of the different types? I know very little about ESDI and IDE (AT) drives and controllers. Thanks, scott coburn brookhaven national laboratory scottc@max.bnl.gov [130.199.128.6] upton, ny, usa
phil@brahms.amd.com (Phil Ngai) (09/20/90)
In article <2132@bnlux0.bnl.gov> scottc@bnlux0.bnl.gov (david scott coburn) writes: |In other words, why should the data transfer rates for a RLL disk with a |host-resident RLL controller differ from a RLL disk with an embedded |SCSI controller? It wouldn't, assuming it was a SCSI RLL system and ignoring caches. But many SCSI system are not simply MFM or RLL. A popular technique these days is ZBR (zone bit recording) where the data per track actually increases as you move from the inner to the outer tracks. Such "geometries" would hopeless confuse DOS but SCSI can hide it. Another trick is to increase the RPMs, typically from 3600 to 5400. (ever calculate the rotational latency at 3600 RPM?) Finally, ESDI systems range from 10 Mbits/sec to 15 and up. A SCSI system coupled to such a fast disk could easily outperform MFM or RLL. If you have the bucks. -- Phil Ngai, phil@amd.com {uunet,decwrl,ucbvax}!amdcad!phil Freedom is dead, long live privacy!
davidsen@sixhub.UUCP (Wm E. Davidsen Jr) (09/23/90)
Sanity check here!
The vendors can claim anything they want for these disks, but in truth
this is how you get the transfer rate: multiply the physical sectors per
track by the rpm of the disk, times 68.2666 and that's how fast it's
coming off (or going on) the disk.
Before you ask:
RPM/60 = rps * 512 = bytes/sec * 8 = bps
512 * 8 / 60 = 68.2666
So your big bucks ESDI drive with 15mbit interface has 34 sectors and
turns at 3600, or ~8mbit. Hummmm. The only drive I know which really has
transfer rate is a 60 spt 2 in drive running 5400 rpm (50% faster) which
goes ~21mbit. And the only place you see those is in trade shows right
now.
What makes a fast disk is (a) seek time, (b) interleave (the
controller can help). Correct setup helps, too. If you can adjust skew
for head to head and track to track you can save almost a full
revolution when you change.
An RLL system 1:1 is about as fast as an ESDI 1:1 in actual practice.
SCSI depends on the controller and software, and MFM is slower however
hacked. There is a *lot* more diference between controllers (and BIOS if
you run DOS) then the interface typically makes.
Hope this blew a little light into the fog.
--
bill davidsen - davidsen@sixhub.uucp (uunet!crdgw1!sixhub!davidsen)
sysop *IX BBS and Public Access UNIX
moderator of comp.binaries.ibm.pc and 80386 mailing list
"Stupidity, like virtue, is its own reward" -me