Re: OSA-Express2 Device Addresses

New Message Reply Date view Thread view Subject view Author view Other groups

Subject: Re: OSA-Express2 Device Addresses
From: Chris Mason (chrismason@belgacom.net)
Date: Sat Aug 29 2009 - 12:06:32 EDT


Tom

Take a look at this redpaper for preliminary information regarding VSWITCH.

Linux on IBM eServer zSeries and S/390: VSWITCH and VLAN Features of z/VM 
4.4
http://www.redbooks.ibm.com/abstracts/redp3719.html

This document tells me that zLinux is "fooled" into "thinking" that it is 
using real QDIO. I would very strongly expect that z/OS together with the 
z/OS Communications Server IP component as well as z/VSE together with any 
software providing an IP service would similarly be "fooled" into "thinking" 
they were using real QDIO.

Recall that "fooling" into "thinking" that something *virtual* is *real* is 
the "raison d' tre" of z/VM and all its forebears.

In case you haven't already you may wish to acquaint yourself with the 
following two manuals:

- Open Systems Adapter-Express Customer's Guide and Reference, SA22-7935-10
- OSA-Express Implementation Guide, SG24-5948-05

Here is a pertinent abstract from the redbook:

<quote>

Question: How many devices should you define?

The answer depends on the CHPID type, the number of TCP/IP stacks, the usage 
of
the OSA-Express Network Traffic Analyzer and SNA definitions required. For 
the
number of devices, consider the following:

- Any OSD CHPID requires at least three devices (read, write, and datapath) 
for each
TCP/IP stack. For z/OS, only the first TCP/IP stack requires three devices; 
any
additional TCP/IP stack requires only one device (datapath).

- If you plan to use the OSA-Express Network Traffic Analyzer (OSAENTA) 
function,
you will need one additional data device. This data device is used to 
transmit the
captured trace data to TCP/IP over a QDIO data device.

- Any OSE CHPID requires two devices (read and write) for each TCP/IP. SNA
requires one device.

</quote>

Chris Mason

----- Original Message ----- 
From: "Tom Rae" <tom_rae@shaw.ca>
To: "VSE Discussion List" <vse-l@Lehigh.EDU>
Sent: Saturday, August 29, 2009 5:21 AM
Subject: Re: OSA-Express2 Device Addresses


> Thanks, Mark.
>
> Initial input that I received suggested that z/VSE did not support virtual 
> NICs as required to couple to a VSWITCH. I will look into this further, 
> though, as a VSWITCH does sound more appealing. CP DEFINE NIC is supposed 
> to create "OSA-type" devices, which z/VSE may recognize as actual OSA 
> devices.
>
> The exclusion of xx00-xx02 and xxFE below was just due to the existing OSA 
> devices at these addresses... no black magic involved!
>
> Mark Post wrote:
>> -snip-
>>
>>> One of the customers that I support is using vCTCAs to connect multiple 
>>> z/VSE and Linux guests to their z/VM TCP/IP and SNA SVMs. External 
>>> connections are provided by QDIO and non-QDIO OSA-Express2 devices (both 
>>> Gb and 1000Base-T) that are attached to the SVMs. We are looking at 
>>> providing more efficient, direct network access to the guests by 
>>> defining additional devices on the OSAs and attaching them directly to 
>>> the guests.
>>>
>>
>> Why not use a VSWITCH instead?  Far less systems programmer effort needed 
>> to maintain, automatic failover is provided by the VSWITCH controller, 
>> you can do Link Aggregation if needed, etc.
>>
>>
>>> and that all of the new devices must be in the xx03-xxFD range.
>>>
>>
>> I have never heard of this before.  Unless you're saying that 00-02 and 
>> 0FE-0FF are already in use, but that would be specific to your situation, 
>> not a generic requirement.
>>
>> 


New Message Reply Date view Thread view Subject view Author view Other groups

This archive was generated by hypermail 2b25 : Wed Sep 02 2009 - 20:20:12 EDT