The IBM z/VM Operating System <IBMVM@listserv.uark.edu> wrote on 04/28/2011 12:19:53 PM:
> From: Alan Ackerman <alan.acker...@bankofamerica.com> > To: IBMVM@listserv.uark.edu > Date: 04/28/2011 12:20 PM > Subject: Restrictions in FCP point-to-point topology > Sent by: The IBM z/VM Operating System <IBMVM@listserv.uark.edu> > > Restrictions in FCP point-to-point topology > > We are planning to use FCP disk with a z/Linux under z/VM system, > running under z/VM 5.4. To reduce cost, we are planning to use the > point-to-point topology. > > Are there any restriction in using this topology? > > I seem to remember some restrictions mentioned in the presentation given > at SHARE when the point-to-point support first appeared in z/VM. > > The load (write) volume will be high (the higher the better), but only > occasionally and for less than a day at a time. Another reason to use > point-to-point is that we think it will perform better without the > switch. > > We want to have a 4TB LVM, spread across 8 LUNs. > > Picture: > > +-------------------+ +-------------------+ > | LPAR 1 | | LPAR 2 | > | +---------------+ | +---------------+ > | | z/VM 1 |-FCP--4TB---FCP-| | z/VM 2 | > | | +-----------+ disk | | +-----------+ > | | | Linux 1 | | | | Linux 2 | > +---+---+-----------+ +---+---+-----------+ > > A wider picture is > > z/OS<--Hipersocket-->Linux-->FCP<--Hipersocket-->z/OS > DB2 disk > > I found a presentation "Introduction to SCSI over FCP for Linux on > System z" at <http://www.vm.ibm.com/education/lvc/lvc1020c.pdf>. It > mentions (on foil 6) the two topologies, point-to-point and switched > fabric, but gives no restrictions. > > Thanks! > > Alan Ackerman > Alan (dot) Ackerman (at) Bank of America (dot) com > Alan, Here are some things that I can think of at the moment: * NPIV is not supported point-to-point, which impacts lun masking (host connections). * Which storage array do you plan on using? Some arrays do not support point-to-point at all, or they support the poor man's arbitrated loop topology with exactly 2 l_ports. Our system z channels support the official FC-FS point-to-point topology. We do not support any arbitrated loop topologies. * We have not measured switched perf vs point-to-point perf here in Poughkeepsie. I think the latency of the switch (probably microseconds) will be much less than the latency of real disks (milliseconds). Especially with lots of writes. I don't think you'll notice the difference, but that is just my guess. Regards, Ray Higgs System z FCP Firmware Development Bld. 706, B42 2455 South Road Poughkeepsie, NY 12601 (845) 435-8666, T/L 295-8666 rayhi...@us.ibm.com