Question about IDE disks
JF Mezei
With proprietary disks such as RA and DSSI, VMS had very good control
and more importantly notification of errors which would then show up in
the SHOW DEVICE output.
There was a phase in time where Alphas supported el-cheapo (made for
Windows PC) IDE (now called ATA) disks. (models such as DS10/DS10L
for instance).
Do those drives (and IDE protocol) give VMS the same level of
notification as previous generation of disk protocols ? Or do they
perform their own internal disk maintenance without bothering the OS ?
In other words, if an IDE disk is starting to make weird noises, does an
error count of 0 on the SHOW DEV provide some indication that the noises
have not resulted in problems with actual disk operation, or could the
disk already be seriously sick and just not reporting its problems to
the OS ?
Richard B. Gilbert
I've had IDE disks running almost continuously for several years. While
failure is possible it is not very common. IDE is only the interface
electronics. I think that virtually all disks have the same sort of
enclosure, the same sort of platters, the same sort of heads. What
differentiates them is: capacity, performance, reliability, and
interface with the system.
Michael Moroney
>I've had IDE disks running almost continuously for several years. While
>failure is possible it is not very common. IDE is only the interface
>electronics. I think that virtually all disks have the same sort of
>enclosure, the same sort of platters, the same sort of heads. What
>differentiates them is: capacity, performance, reliability, and
>interface with the system.
That is true, however it should be pointed out that some drives were made
with lower quality HDAs, and these were almost exclusively IDE for cheap
home PCs. This is why some people say IDE drives are lower qulaity, it's
really just the low quality HDAs weren't ever made into SCSI or whatever
drives.
JF Mezei
that IDE protocol is able to send error messages to the OS, but that the
DQDRIVER on VMS doesn't do anything with them.
However, there appears to be at least an exception to this since the
DQDRIVE is able to log an error when the system is being configured and
it tries very hard to create a drive that doesn't exist. (if you don't
have 2 drives in the system).
What I have done in the mean time is accelerate the moving of data to
the new server. I had planned to do this on a application by application
basis but am now using ZIP to take a snapshot of each top dircetory on
the drive. *(I was told VMS can't handle files greater than 4 gigs on a
remote NFS server, so breaking up the drive into multiple zip files
would alleviate this).
H Vlems
JF,
from the way Seagate issues model numbers, I'd assume that the HDA and
the drive electronics are two distinct products.
Drive electronics (again, just my guess) are part drive specific and
part depending on the connecting bus.
A Seagate model looks like this:
ST <geometry><capacity><connection type>
Where:
<geometry> is a single digit that tells you about the width and
heigth of a drive. E.g. 1 is a 3.5" drive disk, approx. 1.5"high and 3
is a 1"high drive
<capacity> is either in kB or in MB, depending on the age of the
drive in four or five digits
<connection> one or two letters specify the interface. E.g. N is 50
pin SCSI-2, A is IDE and WC is SCSI-SCA
So an ST32550WC is a 3.5" disk, 1" high with a storage capacity of 2.5
GB (formatted, VMS will see approx. 2.1 GB nett storage).
Now many drives were available as families. Example the ST34501 came
in 5 different interface versions N/W/WC/WD/DC.
Which led me to believe that Seagate designs an HDA and slaps a pcb on
it fitted with a specific interface.
Now in this example (The ST34501 is from 1997) the "A" is missing.
Until a couple of years ago SCSI drives had larger capacity than IDE
drives and very likely their HDA's were built and designed for 24x7
usage.
Apparently that difference no longer mattered for the production of
the HDA, hardware MTBF figures just went up, and an HDA was used for
SCSI and IDE.
Today SCSI lags behind, 300 GB is the maximum capacity for SCSI while
IDE is available at 500 GB (perhaps even bigger).
I do not know about the differences between serial SCSI and serial ATA
(SAS and SATA) but I'm willing to propose that these drives share the
same HDA's.
IDE drives (parallel ATA) have a facility called SMART which allowed
the drive to inform the OS about the shape it was in. I don't know
whether VMS can interpret the data but it surely is available. DEC
provided utilities to extract data from DSSI and SCSI devices so
perhaps it is possible to write a similar tool to inspect IDE drive
logic?
Hans
glen herrmannsfeldt
(snip)
> from the way Seagate issues model numbers, I'd assume that the HDA and
> the drive electronics are two distinct products.
> Drive electronics (again, just my guess) are part drive specific and
> part depending on the connecting bus.
> A Seagate model looks like this:
> ST <geometry><capacity><connection type>
Most of the time probably true, but I know of some cases where
matching disk size was a completely different HDA. The cases
that I know of are from a long time ago, though.
> Where:
> <geometry> is a single digit that tells you about the width and
> heigth of a drive. E.g. 1 is a 3.5" drive disk, approx. 1.5"high and 3
> is a 1"high drive
> <capacity> is either in kB or in MB, depending on the age of the
> drive in four or five digits
(snip)
> Apparently that difference no longer mattered for the production of
> the HDA, hardware MTBF figures just went up, and an HDA was used for
> SCSI and IDE.
> Today SCSI lags behind, 300 GB is the maximum capacity for SCSI while
> IDE is available at 500 GB (perhaps even bigger).
> I do not know about the differences between serial SCSI and serial ATA
> (SAS and SATA) but I'm willing to propose that these drives share the
> same HDA's.
One reason is that SCSI drives often spin faster. For a given number
of tracks and bit clock, a faster spinning disk has lower capacity.
Most important, it also has lower latency. (Delay for the disk to
rotate to the desired block.)
> IDE drives (parallel ATA) have a facility called SMART which allowed
> the drive to inform the OS about the shape it was in. I don't know
> whether VMS can interpret the data but it surely is available. DEC
> provided utilities to extract data from DSSI and SCSI devices so
> perhaps it is possible to write a similar tool to inspect IDE drive
> logic?
Even more interesting, ATAPI used for CD-R and DVD-R sends SCSI
commands over the IDE bus. SATA disks, it seems, also accept
SCSI commands and are often configure to the OS as SCSI disks.
(In linux they have sd names instead of hd names.)
-- glen
Michael Moroney
>BTW, a very knowledgeable person has contacted me privately. It appears
>that IDE protocol is able to send error messages to the OS, but that the
>DQDRIVER on VMS doesn't do anything with them.
Judging from the speed of DQDRIVER on Alpha, I assume they also neglected
to put the drive in a mode faster than some default low speed mode.
>However, there appears to be at least an exception to this since the
>DQDRIVE is able to log an error when the system is being configured and
>it tries very hard to create a drive that doesn't exist. (if you don't
>have 2 drives in the system).
DQDRIVER also doesn't do the right thing here, by creating nonfunctional
drives (UCBs) in the IO database for nonexistent drives. DKDRIVER doesn't
create a nonfunctional DKA100:/DKA200: etc. if the only drive on the PKA:
bus is DKA0:.
(I vaguely remember someone mentioning a reason for that behavior, but also
vaguely remember it as a not very good excuse)