>>> On Fri, Apr 25, 2008 at 7:58 AM, in message <[EMAIL PROTECTED]>, "Gentry, Stephen" <[EMAIL PROTECTED]> wrote: -snip- > In a previous IFL/Linux/POC, we set up a Linux running UDB. We wanted to > allow access to that Linux from two sources: existing mainframe apps. > and apps. running on WAS (IBM web app. server). I set up a virtual lan > (hiper-sockets and such) for the apps on the mainframe to get to UDB. > For the WAS part, I used 3 addresses from an OSA card, DEDICATED those > to the Linux/UDB guest. This allowed WAS to avoid going through the VM > TCPIP stack and through the virtual lan to get to UDB, in essence > reducing the length of the path taken to get to UDB.
Network traffic on a VSWITCH doesn't go through z/VM's TCP/IP stack, just CP. (Unless you deliberately configure your virtual network to do so.) The main reason you have a TCP/IP "controller" is to handle the actual OSA hardware. TCP/IP already had all that code built in, so there wasn't a good reason to duplicate it or rewrite it just to manage the hardware the VSWITCH needs to get "outside." Mark Post
