How about etherchannel? Quoting from the AIX 4.3.3 Release Notes,
"Etherchannel is an aggregation technology that ships with AIX 4.3.3. With
Etherchannel, you can produce a single large pipe by combining the
bandwidth of multiple ethernet adapters. The aggregated pipe appears to be
a single ethernet interface to the upper layers (e.g., IP). Each ethernet
adapter shares a single MAC address. There is a single ethernet interface
and a single IP address for the aggregated pipe. For more information about
Etherchannel, see the /usr/lpp/bos/README file."

So if your server and client are both AIX it sounds like it could be done
today. Other OS's may have or be getting similar capabilities.

I only saw the headline and have not researched the details to find the
gotcha's. Even if it would work in your environment, you would still need
to consider other throughput issues, ie how fast you could read and write
the data and how may tape devices you could write to at the same time.

Steve Branch
Phillips Petroleum
e-mail: swbranc AT ppco DOT com




Darrell Kundrik <Darrell_Kundrik AT TRANSALTA DOT COM>@VM.MARIST.EDU> on
12/07/2000 09:25:23 AM

Any replies will be addressed to: "ADSM: Dist Stor Manager"
      <ADSM-L AT VM.MARIST DOT EDU>

Sent by:  "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>


To:   ADSM-L AT VM.MARIST DOT EDU

cc:


Subject:  Two ethernets for redundancy/performance?


We are running NSM here with a gigabit ethernet card on a private network
to back up clients. We would like to install a second gigabit card for
extra throughput and failover in case one card should happen to die. In the
initSID.utl or the dsm.opt files, it looks like I can set:

MAX_SESSIONS        4         #we have 4 tape drives

SERVER              server_a
SESSIONS            2
TCPSERVERADDRESS    xxx.xxx.xxx.xxx      #address of 1st card

SERVER              server_a
SESSIONS            2
TCPSERVERADDRESS    yyy.yyy.yyy.yyy      #address of 2nd card

This should give increase bandwidth and a certain amount of failover. Is
there any method to achieve SESSIONS=4 if one of the cards should fail?

Thanks,
Darrell Kundrik