This note describes the status of DLT data production for the U.S. MACRO
institutions. It is time for the U.S groups to provide their own DLT tapes for
this production. Details are given below.
Test of MACRO DLT tapes in the U.S.
-----------------------------------
At Caltech, we have had a 20 GB (native capacity) DLT drive in operation for a
little over a month now. It appears to perform very well on our RISC machines
running AIX. After some fiddling with setup parameters, it is possible to
read/write to the drive at a rate of about 1.4 MB/s on a local machine. MACRO
data tapes (10 GB) have been read successfully. Also, tape-to-tape copies have
been made from MACRO exabyte data tapes onto DLT at the full speed of operation
of the exabyte drive (around 300 KB/s). System backups are now done using the
DLT at Caltech with a significant improvement in speed and convenience. Chris
Walter has been the tester/user for MACRO and plans to write a note about the
technical performance and tuning. The Caltech drive was purchased from Nth
Generation Computing in San Diego for $5300. Nominally, the internal drive is
built by Quantum. However, the *only* manual which came with the drive was a
Xerox copy of a DEC manual! This makes us suspect that in fact Digital is
really the only company in the business but they launder the drives through
subsidiaries in order to make it look public. (Does anyone know if Quantum is
actually owned by DEC?) Tapes also appear to look identical between the "DEC"
ones at the Gran Sasso and those which we bought from Nth Generation for
Caltech. Hence, I think that we can probably expect no compatibility problems.
At Texas A&M, a 20 GB (native capacity) drive was purchased from Andataco
(again with a "Quantum" internal drive) for $4700. (They got the same cheesy
xeroxed manual.) A MACRO data tape has been read successfully with full speed
data transfer on a local machine. Ash tells me that their machine is an Alpha
running Ultrix and in order to make the drive work he had to get a system patch
off of the network but once that was done it seems to work very well.
I don't know if other U.S. institutions have acquired their drives yet.
However, these experiences seem to show that the technology works on our
systems and with MACRO data with no problem in using the 10 GB tapes from
the Gran Sasso in our 20 GB drives.
Status of DLT production at the Gran Sasso
------------------------------------------
Copies of full MACRO data have been produced for each U.S. institution
as of 1-Dec-1995 (a total of 6 copies) on 10 GB capacity DLT. The ability
to make direct tape-to-tape copies from DLT to DLT on a local machine
has been demonstrated to take only about 2 hours. This means that it will
be relatively easy to go back and fill in missing data from the past.
For those who don't care about getting data now, it could also permit
the option of delaying your production until 20 GB capability is ready.
The laboratory has responded to our request for production on 20 GB tapes
for the U.S. institutions by ordering a 20 GB drive and with a plan to
upgrade one of the 7-tape stackers to a 20 GB drive. The expectation is that
within 5 months we should have the 20 GB production going. Tests on
recording data using the hardware compression available on the drives show
a compression factor of only 20%. Since using compression requires that
some significant buffer be left on each tape (there is apparently no readback
of how much data has actually been written) it is likely easiest to just
fill the tapes as full as possible with no compression. This will also make
it easier to understand the necessary disk space for copying data from tape.
Need of Tapes for Continued Production
--------------------------------------
The production thus far has been using tapes generously supplied by the
laboratory. However, the time has come that the U.S. groups are expected to
supply our own tapes in order to continue production. Caltech has supplied our
first 20 tapes from which I am willing to make loans to other institutions, for
now. However, as in the past, each institution is expected to provide its own
tapes for production. Because with the DLT, tape-to-tape copies are relatively
easy, I suggest two options for U.S. groups to supply tapes:
1. Supply enough 10 GB tapes now for 5-6 months of production (25 tapes should
be sufficient). This will be of particular interest to people who really
want to look at data now.
2. Wait until the 20 GB production is ready (nominally about 5 months) and then
supply only 20 GB tapes when the time comes.
I don't think there are any big factors to recommend one of these options
over the other except for the issue of wanting to look at data now.
At present, the cost for tapes in the U.S. is:
DLT III (10 GB) $45 in quantities less than 100 (as purchased by
Caltech)
$35-$40 for "large" quantities
DLT IV (20 GB) $125 in quantities less than 100 (as purchased by
Caltech)
$90 in quantities less than 100 (recently quoted to
Texas A&M)
$70 for "large" quantities (recently quoted to
Texas A&M)
Hence, although there was a momentary "premium" in the per-byte cost for the
20 GB tapes, according to the recent quotes to Texas A&M this situation
has now rectified itself. Since prices are only falling, I suggest delaying
tape purchase for as long as possible. Our plan at Caltech is to buy only
the necessary 10 GB tapes for the next few months now and then buy 20 GB
tapes when it looks like the production capability is ready.
DEADLINE for Providing Tapes for Continuing Production
------------------------------------------------------
If you want to exercise option 2 for data production, you don't need to send
any tapes to the Gran Sasso very soon. I will send a message at a later time
giving approximately a 1 month warning on providing tapes. However, if you wish
to exercise option 1, you need to provide 10 GB tapes at the Gran Sasso as soon
as possible. The supply of tapes that I brought from Caltech will allow 6-copy
production to continue for 3 weeks or so but then it runs out. Since we are
interested in looking at the data now at Caltech we will start overwriting
tapes in order to ensure our own production. Hence, for option 1 you should act
quickly to purchase some tapes (ours were delivered 2 days after placing the
order) and get them to somebody coming to Italy. The next U.S. shiftworker
scheduled to come to Italy is someone from Boston. I will also be making a trip
from the U.S. to Italy in mid-February. If you know that you will pursue
option 2 for your institution, please inform me now so that we don't have
to produce the 10 GB tapes for you.
If tapes for your institution are not at the Gran Sasso by February 15th,
it is assumed that you want option 2 for data distribution.