rpburry@ncs.dnd.ca (Paul Burry) (03/16/91)
Here's a few questions regarding SCSI and IPI.
1) What are the advantages of each of these disk interfaces.
2) What are the disadvantages of each of these disk interfaces.
3) Is SCSI-2 really that much slower than IPI-2?
4) Are there any drives and Sun compatible controllers that take advantage of
the extra features of SCSI-2 (ie. 16 or 32 bits wide, 10mb/s transfer)
5) Has anyone actually seen a fast IPI controller and disk work together
on a Sun 4/490?
6) Xylogics claims that their SV6892 will give a 24 MB/second transfer rate
when used with Seagate's ST2105 : ref. UnixWorld p136. Are these shipping
now? Has anyone had this combination work on a 4/490?
--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Paul Burry
Voice: (613)-991-7325 Internet: rpburry@ncs.dnd.ca
Fax: (613)-991-7323 UUCP: ..!{uunet,cunews}!ncs.dnd.ca!rpburrykaufman@neon.Stanford.EDU (Marc T. Kaufman) (03/16/91)
In article <1991Mar16.012045.20207@ncs.dnd.ca> rpburry@ncs.dnd.ca (Paul Burry) writes: >6) Xylogics claims that their SV6892 will give a 24 MB/second transfer rate > when used with Seagate's ST2105 : ref. UnixWorld p136. Are these shipping > now? Has anyone had this combination work on a 4/490? What's an ST2105? - my Seagate catalog (which I just got this week) does not list it. For reference, the raw data rate of an ST2106 (106MB) is 10 Mbits/sec (1.25 MB/sec). How do they get 24 MB (192 Mbits)? Marc Kaufman (kaufman@Neon.stanford.edu)
rpburry@ncs.dnd.ca (Paul Burry) (03/17/91)
In article <1991Mar16.072821.240@neon.Stanford.EDU> kaufman@neon.Stanford.EDU (Marc T. Kaufman) writes: |In article <1991Mar16.012045.20207@ncs.dnd.ca> rpburry@ncs.dnd.ca (Paul Burry) writes: | |>6) Xylogics claims that their SV6892 will give a 24 MB/second transfer rate |> when used with Seagate's ST2105 : ref. UnixWorld p136. Are these shipping |> now? Has anyone had this combination work on a 4/490? | |What's an ST2105? - my Seagate catalog (which I just got this week) does not |list it. For reference, the raw data rate of an ST2106 (106MB) is |10 Mbits/sec (1.25 MB/sec). How do they get 24 MB (192 Mbits)? | |Marc Kaufman (kaufman@Neon.stanford.edu) I'm afraid that I may have quoted the Seagate model number wrong (my catalog is at work). According to the UnixWorld quote (p136 "IPI in the Sky") "... and currently supports Seagate Technology's Sabre parallel transfer disk. With a storage capacity of 2105 megabytes, the Sabre product uses eight read/write heads simultaneously for transfer rates up to 24 megabytes per second. ..." The catalog also lists a IPI-2 3GB disk drive with a transfer rate of 27MB/second. -- =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= Paul Burry Voice: (613)-991-7325 Internet: rpburry@ncs.dnd.ca Fax: (613)-991-7323 UUCP: ..!{uunet,cunews}!ncs.dnd.ca!rpburry
kaufman@neon.Stanford.EDU (Marc T. Kaufman) (03/18/91)
In article <1991Mar17.145652.27084@ncs.dnd.ca> rpburry@ncs.dnd.ca (Paul Burry) writes: >According to the UnixWorld quote (p136 "IPI in the Sky") >"... and currently supports Seagate Technology's Sabre parallel transfer disk. >With a storage capacity of 2105 megabytes, the Sabre product uses eight >read/write heads simultaneously for transfer rates up to 24 megabytes >per second. ..." That's the ST82105 Sabre-6 8HP, 8" drive with IPI-2 interface. It has 16 R/W heads. >The catalog also lists a IPI-2 3GB disk drive with a transfer rate of >27MB/second. That's the ST82368K Sabre-6 9HP, 2368 megabytes (unformatted) with 18 R/W heads. The Sabre 7 drives, which are 3000+ megabytes, only list transfer rates of 4.67 and 9.34 MB/sec. Apparently the Sabre-6 drives are organized as 400000 bytes/track, 2 tracks per cylinder, by having half the heads working at once. Marc Kaufman (kaufman@Neon.stanford.edu)
taylor@intellistor.com (Dick Taylor) (03/19/91)
In article <1991Mar16.012045.20207@ncs.dnd.ca> rpburry@ncs.dnd.ca (Paul Burry) writes: >Here's a few questions regarding SCSI and IPI. > >1) What are the advantages of each of these disk interfaces. > >2) What are the disadvantages of each of these disk interfaces. > >3) Is SCSI-2 really that much slower than IPI-2? Actually, depending on the application, SCSI-2 can be significantly "faster" (take that as sustained throughput or burst rate or whatever metric you want) than IPI-2. Or not. All three questions above are apples vs. oranges questions. SCSI and SCSI-2 are parallel-transfer high-level "intelligent" interfaces which allow you to mix a wide variety of peripherals on a single bus. IPI-2 is a device-level interface which currently has higher actual device transfer rates but which is not by itself a sufficient I/O bus for a system (that is, you still have to have something for tapes and optical and the like). Actual comparisons between the two are so viciously application-specific that it's hard to credit any general argument. Not that folks haven't tried, but it seems like the more general the statement is, the less likely it is to be true in any specific instance. > >4) Are there any drives and Sun compatible controllers that take advantage of > the extra features of SCSI-2 (ie. 16 or 32 bits wide, 10mb/s transfer) > Silicon to support fast SCSI is becoming available. It's gonna be fun to see if everyone can get the cabling, noise, and termination issues worked out. Wide transfers were a victim of the standardization process (I'm still unclear on how that worked out) and aren't readily available except for a few products like Ciprico's companion HBA for their array controller. (And I don't know the exact state of that one -- Ciprico and I parted company a while back.) >6) Xylogics claims that their SV6892 will give a 24 MB/second transfer rate > when used with Seagate's ST2105 : ref. UnixWorld p136. Are these shipping > now? Has anyone had this combination work on a 4/490? > I can believe that Xylogics will show that transfer rate (peak) through their controller. I can't believe for a moment that you'll see speeds anything like that in a 4/490. The bottlenecks shift -- you suddenly have to worry about the speed of Sun's VMEbus (it's not a real barn-burner), memory bandwidth, and system issues (like the fact that there's no WAY that the pagedaemon can keep up freeing pages at that rate). Disks are rapidly getting fast enough that some serious system work will be needed to keep pace on the workstation side. Not that it's Sun's fault that they haven't kept pace -- this is one of those things where it's a race between the system and the disk, and disks are, for the moment, catching up to currently-used technology. > > >-- >=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= >Paul Burry >Voice: (613)-991-7325 Internet: rpburry@ncs.dnd.ca >Fax: (613)-991-7323 UUCP: ..!{uunet,cunews}!ncs.dnd.ca!rpburry