On Fri, Nov 08, 2002 at 08:44:20AM -0600, Nix, Robert P. wrote:
> Given an IFL running zVM and several Linux/390 images, is it better to
> fan out to all the Linux images from zVM's TCPIP, or should TCPIP talk to a
> selection of images, with these images each handling several end
> machines, more like a tree structure? What would be the advantages and
> disadvantages of either method, and is there a "break-even" point
> below which you'd want to fan from TCPIP, but above which you'd want
> "helper" images?
I'd say that the answer is "none of the above."
Which version of z/VM? If at all possible, use a guest LAN. It makes
your life a *whole* lot simpler. If you can't, I'd say use about six
downstream Linux images per upstream router; IUCV is theoretically
faster, but I have found CTC somewhat easier to configure.
Now, you're using SuSE, so that may be a stumbling block too. IIRC, the
totally-free version (beer, not speech, for those of you keeping score
at home) of SuSE doesn't do HiperSockets. But if you have either the
$500 trial or a support contract then you have access to the service
releases, which do let you use HiperSockets. And if you don't, then
(IMHO) you shouldn't be using SuSE--if you're going to be your own
support, you may as well be your own support with a less antiquated
distribution. I'm going to surprise exactly no one by saying here,
"Debian." Largely because I haven't played with RH in a long time, and
I know that Debian works just fine (albeit taintedly) with the qeth
drivers and guest LANs. To wit:
debian:~# uname -a
Linux debian 2.4.19 #1 SMP Thu Nov 7 13:33:12 CST 2002 s390 unknown
debian:~# ifconfig hsi0
hsi0 Link encap:Ethernet HWaddr 00:00:00:00:00:00
inet addr:192.168.129.7 Mask:255.255.255.0
inet6 addr: fe80::200:ff:fe00:0/10 Scope:Link
UP RUNNING NOARP MULTICAST MTU:8192 Metric:1
RX packets:228 errors:0 dropped:0 overruns:0 frame:0
TX packets:143 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:100
RX bytes:20515 (20.0 KiB) TX bytes:22262 (21.7 KiB)
Interrupt:15
debian:~# route -n
Kernel IP routing table
Destination Gateway Genmask Flags Metric Ref Use Iface
192.168.131.65 0.0.0.0 255.255.255.255 UH 0 0 0 ctc0
192.168.129.0 0.0.0.0 255.255.255.0 U 0 0 0 hsi0
0.0.0.0 192.168.129.1 0.0.0.0 UG 0 0 0 hsi0
debian:~#
You'll notice that I'm routing through the HiperSockets connection, not
the ctc.
Adam
