tundra@ux1.cso.uiuc.edu (John Kemp) (03/14/90)
Here is the dilemma. We want to have a discless 340
booting off of a 835 (because it is a cheap way to get
a SCSI port). And we want to use a
* 20 Gbyte r/w Optical JukeBox *
drive attached to the 340 and access it from the 835.
Can it be done? And if so, what are the limitations.
Someone has told us that if the 340 is discless you CAN NOT
do block I/O to the Optical drive, but that if the 340 has
it's own disk, you CAN do block I/O to the Optical. Does
anybody understand the reasoning behind this? If we just mounted
every platter via NFS could we avoid this?
Suffering the pains of watching the world
migrate from HP-IB to SCSI...
-------- john kemp ( ( )_ internet - kemp@uiatma.atmos.uiuc.edu
----- ( ( __) decnet - uiatmb::kemp
--- univ of illinois (_ ( __) bitnet - {uunet,convex}
-- dept of atmos sci .(____). !uiucuxc!uiatma!kemp
- 105 s gregory ave ... phone - (217) 333-6881
- urbana, il 61801 ... fax - (217) 444-4393
--
-------- john kemp ( ( )_ internet - kemp@uiatma.atmos.uiuc.edu
----- ( ( __) decnet - uiatmb::kemp
--- univ of illinois (_ ( __) bitnet - {uunet,convex}rjn@hpfcso.HP.COM (Bob Niland) (03/15/90)
re: > We want to have a discless 340 booting off of a 835 (because it is a > cheap way to get a SCSI port). And we want to use a > * 20 Gbyte r/w Optical JukeBox * Can't be done, yet. Diskless clients can't have locally mounted file systems. They can have swap disk, raw device access, but no block-I/O. For the moment, the 340 will have to be self-hosted. The most economical way to do that is to get the 200 Mbyte internal disk (available as an upgrade if you already own the 340). Then you can NFS mount the JukeBox. At HP-UX 8.0, there is a good chance that we will support local mounts, but then the Series 800 will have native SCSI support at 8.0, eliminating the need for the 340-size "lump" in the cable between the 800 and Jukebox. > Suffering the pains of watching the world migrate from HP-IB to SCSI... So are we. Regards, Hewlett-Packard Bob Niland 3404 East Harmony Road rjn%hpfcrjn@hplabs.HP.COM Fort Collins UUCP: [hplabs|hpu*!hpfcse]!hpfcla!rjn CO 80525-9599 This response does not represent the official position of, or statement by, the Hewlett-Packard Company. The above data is provided for informational purposes only. It is supplied without warranty of any kind.
tsuia@spock (Alan Tsui) (03/20/90)
In article <7370103@hpfcso.HP.COM> rjn@hpfcso.HP.COM (Bob Niland) writes: >Can't be done, yet. Diskless clients can't have locally mounted file >systems. They can have swap disk, raw device access, but no block-I/O. For Is it really that hard to do? Sun can do that 12 months ago! Would that be a marketing issue or technical issue? >> Suffering the pains of watching the world migrate from HP-IB to SCSI... Well, if ther are 2GB HP-IB disk drive with faster read/write times and cheaper in price than SCSI, I don't see why hp customers would ever want to migrate from HP-IB to SCSI. -- Alan Tsui (613) 592-2122 Kanata, Ontario, Canada ...!uunet!mitel!tsui
rjn@hpfcso.HP.COM (Bob Niland) (03/22/90)
re: >>Can't be done, yet. Diskless clients can't have locally mounted file >>systems. They can have swap disk, raw device access, but no block-I/O. > Is it really that hard to do? There are all sorts of thorny technical problems, not the least of which is defining the capability set.... do we allow both swap & fs on the disk? Can it/they be shared between cnodes/server? What happens if a cnode crashes, etc. Nonetheless, local mount is planned for 8.0 (but this is not a firm commitment by HP that it will be there). > Would that be a marketing issue or technical issue? Marketing has wanted it since 6.0. It has always been a technical issue. >> Suffering the pains of watching the world migrate from HP-IB to SCSI... > Well, if ther are 2GB HP-IB disk drive with faster read/write times > and cheaper in price than SCSI, I don't see why hp customers would > ever want to migrate from HP-IB to SCSI. Our SCSI disks are already 2x the file-system throughput of HP-IB, and with the advent of SCSI-2, HP-IB performance will be left in the dust before too long. Future HP-IB disks promise to be more expensive too, since the mechs now always have controllers that are industry standard (SCSI, ESDI), and thus an external bridge-board (usually ESDI-to-HP-IB) is required. Regards, Hewlett-Packard Bob Niland rjn%hpfcrjn@hplabs.HP.COM 3404 East Harmony Road UUCP: [hplabs|hpfcse]!hpfcla!rjn Ft Collins CO 80525-9599
rml@hpfcdc.HP.COM (Bob Lenk) (03/28/90)
In article <2600@quirkc> tsuia@spock (Alan Tsui) writes: > >Can't be done, yet. Diskless clients can't have locally mounted file > >systems. They can have swap disk, raw device access, but no block-I/O. > > Is it really that hard to do? Sun can do that 12 months ago! > Would that be a marketing issue or technical issue? SunOS and HP-UX have a very different design for diskless. HP-UX is designed around a cluster of machines sharing a common file system hierarchy. One of the advantages is that the entire cluster is administered largely as a single system. SunOS is designed as individual machines each mounting a set of file systems (including the root file system) over the network. One advantage is that file systems local to each machine fit more simply into this model. Supporting local file systems is not *that* hard, but it is not trivial. The choice to work on other things instead can be considered a marketing choice. Bob Lenk rml@hpfcla.hp.com hplabs!hpfcla!rml