FCIP replication behaviour

[chris_shaw]

Hi,

I'm currently testing replication of IBM SVC with Cisco 9222i switches. This is using the Global (asynchronous) mirror option. Both Production and DR storage is at the Production site at present

When I throttle the bandwidth with the tcp-max-bandwidth on the "Production" side we start to get I/O latency on the hosts attaching to the SVC. When we throttle bandwidth at the "DR" side we don't notice a problem.

As this is asynchronous replication the hosts shouldn't be getting I/O latency so the only thing I can think is that when I/O is backing up on the 9222i for transmission it is chewing up switch resources which is slowing the servers access to storage.

Does anyone have any ideas?

Any help appreciated.

Regards

Chris

[jsivulka]

Depending on the array that are used, you can use write acceleration to make the replication much fast. This is a licensed feature though and requires there is only one FCIP tunnel between switches.

http://www-01.ibm.com/common/ssi/cgi-bin/ssialias?subtype=ca&infotype=an&appname=iSource&supplier=897&letternum=ENUS107-622

[ryan_heiserman]

This is unrelated in many ways, however I have a client of mine in Washington looking for support on the Cisco MDS 9222i platform.

If you know anyone that could be of assistance I would appreciate any help.

Thank you,

Ryan C. Heiserman

rheiserman@crescentsolutions.net

[chris_shaw]

Hello Ryan,

I am UK based so unfortunately cannot help with US assistance.

Regards

Chris

[ryan_heiserman]

Chris,

Thank you for getting back to me on this.

If anyone comes to mind feel free to pass along my info.

Thanks again,

Ryan

rheiserman@crescentsolutions.net

[ryan_heiserman]

Chris,

This is unrelated in many ways, however I have a client of mine in Washington looking for support on the Cisco MDS 9222i platform.

If you know anyone that could be of assistance I would appreciate any help.

Thanks Chris,

Ryan C. Heiserman

rheiserman@crescentsolutions.net

[chris_shaw]

We seem to have narrowed down the problem. Their is a Brocade switch in the environment and when we disable the ISL to this switch it seems to negate the problem. One of the other issues we have is that servers plugging into the Brocade have their port disabled when they login to the switch. This problem disappears when we drop the FCIP link to the other site.

So it looks like there is an interoperability issue. IBM are looking into it.

Thanks

Chris