At this stage I do not think this is a snapmirror issue. There are a lot of people out there successfully doing transfers over higher latency links than the 20Mbit I have.

I was unintentionally misleading with the "ftp between filers" statement.

The transfer was actually between 2 unix hosts that had directories mounted on the filers. So there are differences in the way the tcp traffic will be handled. It does establish that the network setup we currently have CAN be saturated. However, it does not establish that we have it setup optimally for the snapmirror we are trying to do.

After talking with some very good NetApp engineers, we decided to do a more specific test limited to filer-filer communication by doing; 1) A snapmirror on the same network segment to an R100. This is underway now using exactly the same volumes as I have mentioned. 2) An NDMPcopy directly to the DR/target filer. This will need to wait unitl a "quiet" time on the WAN so I can hog it and not affect the already slow snapmirror.

The results of these should be able to help us further define whether snapmirror is the issue or not.

A tool suggested by another toasters member is pchar. I am very keen to put it on my network between sites and see what it spits out. Just don't feel like working all weekend!!

Thanks, Aaron

-----Original Message----- From: Todd C. Merrill [mailto:tmerrill@mathworks.com] Sent: Saturday, March 08, 2003 7:18 AM To: Hill, Aaron Subject: Re: Snapmirror throughput question

On Thu, 6 Mar 2003, Hill, Aaron wrote:

6.   ftp transfers have been tested between filers and these saturate
   

the 20Mbit immediately

How do you do this, ftp directly between filers?

The fact that *this* filer-to-filer communication saturates the link seems to be quite damning to snapmirror...

Until next time...

The MathWorks, Inc. 508-647-7000 x7792 3 Apple Hill Drive, Natick, MA 01760-2098 508-647-7001 FAX tmerrill@mathworks.com http://www.mathworks.com ---