Can we configure link based IPMP in private connectivity in Oracle RAC

Hi

I would like to know whether we can configure link based IPMP in private connectivity in Oracle RAC

Regarsd

---------- Post updated at 04:35 PM ---------- Previous update was at 04:27 PM ----------

Here I am taking about in case of private connectivity through cross cable

OK, if you have two nics on server A with crossover cables to two nics on server b, it will be no different than if they were connected with a hub, bridge or switch. They each need their own small IP address nets. Then, why should IPMP care what the network fabric is?

I am not personally experienced with crossover cables. I heard that it can be hard to get them to behave if one server is powered off, but perhaps some nic firmware has stepped up to this. It seems like a frequently overlooked inexpensive and fast network fabric, with no collisions and in full duplex with not much length limitation. NICs are cheap, and wire is far cheaper than gigabit switches! You might run out of slots after a while. There is much need for this sort of bandwidth with high performance SAN and RDBMS client-server configurations.

The following is an active/standby link-based IPMP on Solaris 9 or 10.
Supposed that nic0 has got already an IP address, it loos like:

# cat /etc/hostname.nic0
myhostname

Now do these changes:

# cat /etc/hostname.nic0
myhostname group groupname
# cat /etc/hostname.nic1
group groupname standby

Activation without reboot:

# ifconfig nic0 `cat /etc/hostname.nic0` up
# ifconfig nic1 `cat /etc/hostname.nic1` up

ifconfig -a looks a bit funny, because in.mpathd moves the settings from nic1 to nic0:1
(it will move back to nic1 during a failover).

First of all thanks for the replying to DGPickett & MadeInGermany.

DGPickett -> Are you trying to say that if we have two nics on server A with crossover cables to two nics on server b, then link based IPMP will work between server A & server B. Kindly let me know on this.

Just as easily as it does with any other IP medium. What does it care? It could be PPP or ISDN or hub, and still it is just another IP medium to the same destination. The instructions are how to bond additional IP and names to the path names so communication on the shared IP is sent out on either medium, for both double bandwidth and failover.

Just make sure it is configured with something to watch for outages, as one outage ignored eventually becomes two! RAID5 and the like have this weakness - It starts running on 4 drives but no alert, so nobody ever fixes the fifth drive. A complex medium has many demands n the administrators. Luckily, a cable and two NICs are pretty minimal and reliable. Biggest threat is human tripping on cable!

Remember it is full duplex, so length does not hurt bandwidth, and does very little to latency at a nanosecond a foot or less (speed of light lower in cables due to electromagnetic storage=capacitance and magnetic storage=inductance. The N Ohm rating on the cable says that a light second long cable would draw current like a N ohm resistor for two seconds, and then it depends on what was on the end of the wire. If N ohms load, nothing, else reflected standing wave of voltage (too much resisitance or open) or current (too little resistance or aa short) to destroy transmitter transistors.

1 Like

I have been searching for a coherent explanation why telecom cable are rated in 'ohms' for two decades. Thank you!

Actually, it will take about 3.2 seconds on average coax, because it does not propagate at full light speed.

When at Honeywell, the IRS bought additional one microsecond cycle core memory for their H-2050 from Cambridge Memories. Somehow someone made some parallel interface cables with slower dialectrics or different effective conducter diameters: slow coax! While they did not exceed cable length, the timing was too late. As my friend Jasper Farrington told them, 9 nanoseconds late is still late. Late means broken.

The ARRL Handbook I perused throught my youth ARRL Handbook 2013 had interesting tables on how much to shorten an antenna from the nominal quarter wave length for different conductor diameters and such. Interestingly, a quarter wave whip over a ground plane is 75 ohms like many flavors of coax, and a half wave antenna with the ends connected together is 300 ohms, so you can make a nice antenna from 300 ohm flat lead, shorting the ends and attaching the 300 ohm feed in the middle, or sometimes a little off middle to widen the banwidth for shorter and longer half waves. You see this on a lot of FM sets, where the 100 MHZ into 300 million meters per second gives you a 1.5 meter antenna, but a bit shorter as the propagation is a bit slow. These work on VHF tv as well, as FM is adjacent to channel 6, which many FM radios can hear at the bottom.