You should see a max of about 17.5Mb/sec using vol copy between two Gig interfaces...two 100Bt interfaces..well, then you'll only do 100Mbit.

Vol copy is free.

Its a block for block copy, so the usual NFS performance numbers do not apply as far as how fast you can move data, the bottleneck will be on the destination end. I will preserve snapshots as well.

-----Original Message----- From: Steve Losen [mailto:scl@sasha.acc.virginia.edu] Sent: Thursday, May 25, 2000 8:11 AM To: toasters@mathworks.com Subject: vol copy speed?

We have two F630s in a CF pair, with a total of 7 shelves of 9G drives. We are upgrading to 5.3.5R2P2 and adding 12 36G drives in 2 shelves. We have enough spare slots and spare disks in our existing 7 shelves to depopulate one shelf and leave us with

42 9G disks in 6 shelves

12 36G disks in 2 shelves

We want to do some reshuffling of our shelves between the two filers. We plan to copy the following volume to a new volume consisting of 10 36G disks (leaving 2 hot spares).

Filesystem kbytes used avail capacity Mounted on /vol/vol0/ 139160864 115462812 23698052 83% /vol/vol0/ /vol/vol0/.snapshot 34790216 12362380 22427836 36% /vol/vol0/.snapshot

We are thinking about using vol copy and we would like to preserve all the snapshots.

I think it would be faster to copy locally. However, we have 100Mb ethernet between the two filers, so would it be faster to hook up the shelves on different filers and vol copy across the net?

If this copy can be done in under, say, 5 hours, we have the luxury of turning off NFS and CIFS and giving vol copy 100% of the CPU. We have to shut down both filers for OS upgrade, recabling shelves, etc., so downtime is inevitable. But if this copy is going to take over 5 hours, that's probably too long.

Is there any information about how fast vol copy runs?

Am I correct that vol copy does not require a special license?

We figure that we have two options here

1) one lengthy downtime where we use vol copy during the downtime.

2) two shorter downtimes. In the first downtime we upgrade the OS and add the 36G shelves. Then we bring the filers back up and run a relatively slow dump/restore that loses snapshots. In the second downtime we run an incremental dump/restore and move the disk shelves to their final locations.

If vol copy is "fast enough," option 1 would be easier and less disruptive than option 2.

Steve Losen scl@virginia.edu phone: 804-924-0640

University of Virginia ITC Unix Support