gnu@toad.com (04/11/90)
[For 1.2 GB, 3Mbyte/sec or 6Mbyte/sec transfer rates, 16-18ms average access] > The lowest prices we've heard of for the 97209-1230 and 97229-1150 have > been ~$6800 and ~$11500, respectively. I have found two synchronous SCSI drives in the same performance range, with much lower prices. The Imprimis Wren-7 is a 1.2GB drive with a sub-20ms average access time and up to 4MB/sec synchronous SCSI transfer rate. Hewlett-Packard also makes a 1.2GB SCSI drive with 16ms average access time, 4MByte/sec transfer rate synchronous, 1.5MByte/sec asynchronous. The Imprimis drive (without power supply or cabinet) is available from Arrow Electronics for ~$4000. The HP drive is available from Hybrid Systems (+1 617 357 1838) for $4150. It comes with a 5-year warrantee from HP. Besides the benefit of a 40% price reduction, these are 5-1/4'' drives, taking trivial space and power and not requiring rack mounting. (Hybrid can also sell you the Imprimis drive, and packaging and cables and such.) The SPARCstation-1 supports synchronous SCSI under SunOS 4.0.3c. You have to enable it by patching the kernel flag word "scsi_options"; see /usr/include/scsi/conf/autoconf.h. It is disabled by default because old SCSI cabling is not high enough quality to run sync SCSI, and Sun didn't want customers seeing error when cabling up old shoeboxes and such. (The kernel recovers and turns off sync scsi, but if the failing operation is on a tape drive, it is hard to recover from. Disks just retry it async.) I'm fuzzy on this, but I think the 3/80 has the same SCSI hardware, but its kernel enables sync scsi by default, due to different teams making the decision. None of the Sun VMEbus SCSI host adapters support sync scsi, as far as I know. I have a SCSI-3 adapter and it does not. My HP drive is "in the mail" and I can let you know how it works in a few weeks. I've had nothing but pleasure from my two Imprimis Wren-5's so I expect that a Wren-7 would do fine too. It was the 5-year warrantee from HP that swayed me in their direction. I don't know of any SCSI drives that have hit the 6MB/sec transfer rate of the $11500 IPI drive (the one Sun doesn't sell, though their salesmen talk big about the superior transfer rate of IPI drives over SMD and SCSI). But for that money you could buy TWO 1.2gig SCSI disks and TWO host adapters, for a similar aggregate data rate but twice the access arms, twice the capacity, and the ability to transfer data on one drive while the other is seeking. (Yes, Virginia, you can plug multiple SCSI host adapters on S-bus cards into the SPARCstation-1. I've run into some people who didn't even know it had a bus!) John Gilmore
ekrell@ulysses.att.com (04/11/90)
You can't just look at the transfer rate and seek time to compare 2
different disk subsystems. The controller makes a big difference. Example:
we have a Sun 4/490 and a Solbourne Series 5/802. Both machines have the
Imprimis Sabre 1.2GB disks. Sun uses the disk with the IPI interface and
Solbourne uses the SMD interface. Other than that, the two disks are
identical (both have a 3MB/s transfer rate).
The Sun has 2 of these disks on the Sun IPI controller. The Solbourne has
2 of the SMD disks on the Xylogics 753 controller. Both systems have the
same SPARC CPU and 32MB of memory and are running basically the same
operating system (SunOS 4.0.3).
Running disk I/O benchmarks reveals that the IPI controller is about 25%
faster than the XY753 in writing data to the disk (reading times were
comparable). This is probably due to the 1MB cache on the Sun IPI
controller, but the point is that 2 different controllers with essentially
the same disk can perform quite differently.
Eduardo Krell AT&T Bell Laboratories, Murray Hill, NJ
UUCP: {att,decvax,ucbvax}!ulysses!ekrell Internet: ekrell@ulysses.att.comseeger@manatee.cis.ufl.edu (F. L. Charles Seeger III) (04/12/90)
In article <6551@brazos.Rice.edu> gnu@toad.com writes: |X-Sun-Spots-Digest: Volume 9, Issue 117, message 8 |[For 1.2 GB, 3Mbyte/sec or 6Mbyte/sec transfer rates, 16-18ms average access] | |I have found two synchronous SCSI drives in the same performance range, |with much lower prices. [and mentions "up to 4MB/sec synchronous SCSI transfer rate"] Though modern SCSI drives are quite cost effective, the 4 MB/s transfer rate is misleading because this represents a drive cache to SCSI bus transfer. The raw rate from the media to the cache varies with drives, but figures in the range of 2.5-3.0 MB/s are the best that I have seen. That is a long way from the IPI 6 MB/s mark (which is achieved by reading from two heads in parallel). However, these drives don't make sense everywhere. If the workload is mostly disk service over an Ethernet, then the 6 MB/s can't be fully utilized. And they are rather expensive. But, if you have a multi-user machine, FDDI nets, or disk intensive applications to run locally, then the 2+ speed advantage may be just the ticket. So, I take strong exception to the statement in the subject line that SCSI disks are as fast as the IPI disks. They aren't. Even using a pair of SCSI disks in parallel on separate controllers may not be nearly as fast. It all depends on the application. However, I do agree that in most historic Sun applications, the SCSI drives are a better choice. Afterall, the people at Auspex are pretty bright, and that is their choice. Getting the most reliable drives available is another lesson that we can all learn from them. Seek times and tranfer rates aren't everything. And Sun is not yet delivering the 6 MB/s drives.... But, most of us have IPI and FDDI in our future. Just a matter of time. Charles Seeger E301 CSE Building Work: +1 904 392 1508 CIS Department University of Florida FAX: +1 904 392 1220 seeger@ufl.edu Gainesville, FL 32611 Home: +1 904 375 1819
prc@erbe.se (Robert Claeson) (04/13/90)
In article <6582@brazos.Rice.edu>, ekrell@ulysses.att.com writes:
: The Sun has 2 of these disks on the Sun IPI controller. The Solbourne has
: 2 of the SMD disks on the Xylogics 753 controller. Both systems have the
: same SPARC CPU and 32MB of memory and are running basically the same
: operating system (SunOS 4.0.3).
: Running disk I/O benchmarks reveals that the IPI controller is about 25%
: faster than the XY753 in writing data to the disk (reading times were
: comparable).
Another thing to remember is that 3 MB/sec pushes the limit of an SMD
interface. Sync SCSI can readily to 4 MB/sec, with 10 MB/sec interfaces
coming along (for 8 bit parallel transfers; 40 MB/sec with 32 bit
transfers).
Robert Claeson E-mail: rclaeson@erbe.se
ERBE DATA ABandrew@alice.att.com (04/14/90)
Thinking ahead a year or so, SCSI-2 will be the interface to get. IPI seems like an interim standard that will useful now and for a couple of years but SCSI-2 seems the wave of the future: faster than IPI-2 (and who knows when people will bother to implement IPI-3) and backward compatible with current SCSI.