I have a netapp f760 clustered running 6.01R3 with the following slot assignments:
slot 9 nvram slot 8 cluster interconnect slot 7 redundant fiber loop slot 6 intel fiber gig card slot 5 intel 100mb card slot 4 primary fiber loop
One 7 x 18gb fiber shelf, one volume.
I can't seem to get more than 10MB througput a little worse than on the 100mb card. The cpu numbers drop but so far the gig card isn't a big win.
I've tested from a sun 420 with a syskonnect fiber gig card in the fast pci slot. I've played with about every combo of negotiation, etc. Tested with a cisco 6408 gig card in a cisco 6509 switch and also tested card to card with just a fiber cable.
Here is sysstat:
time dd if=/dev/zero of=./100m.1 count=100000 bs=1024
2% 0 0 0 0 0 32 16 0 0 38 2% 0 0 0 0 0 0 0 0 0 38 2% 0 0 0 0 0 0 0 0 0 38 2% 0 0 0 0 0 32 16 0 0 38 2% 0 0 0 0 0 0 0 0 0 38 88% 1620 0 0 13874 327 496 260 0 0 38 87% 1216 0 0 10429 246 1016 10288 0 0 38 84% 983 0 0 8490 199 741 14470 0 0 38 77% 955 0 0 8190 193 668 13712 0 0 38 73% 956 0 0 8198 193 1121 8725 0 0 38 73% 955 0 0 8190 193 1380 10912 0 0 38 77% 956 0 0 8198 193 1586 10282 0 0 38 81% 1123 0 0 9570 227 488 9646 0 0 38 83% 1133 0 0 9718 229 1049 9810 0 0 38 85% 1224 0 0 10499 247 384 9954 0 0 38 CPU NFS CIFS HTTP Net kB/s Disk kB/s Tape kB/s Cache in out read write read write age 75% 855 0 0 7331 173 1522 12376 0 0 38 74% 534 0 0 4571 108 4268 13680 0 0 38 7% 0 0 0 0 0 116 3844 0 0 38 2% 0 0 0 0 0 32 16 0 0 38 2% 0 0 0 0 0 0 0 0 0 38 2% 0 0 0 0 0 0 0 0 0 38 2% 0 0 0 0 0 32 16 0 0 38 Questions:
1) is it better to have the primary loop in slot 8 and move the cluster interconnect? 2) how is everybody elses netapps laid out? 3) do i need more disk to get better performance?
Hey Art.
Slot assignments can be found at: http://now.netapp.com/NOW/knowledge/docs/hardware/NetApp/syscfg/sysconfig/in dex.htm
Per NetApp: Your primary FCAL loop needs to be in slot 4. The additional loop can be in any available slot so, you've got this correct.
Single port 10/100 card is supposed to go in slots 1,2 or 3.
For Gb-E cards: For optimum system performance, Network Appliance recommends that you install the first Gigabit adapter in the order of slot 8, slot 7, slot 6, and slot 5. If slot 8 is occupied, then use slot 7; and if slots 8, 7, and 6 are occupied, use slot 5, etc...
You specifically mention Intel NICs. Even though NetApp OEMs Intel NICs, you have to use the NetApp versions. These are NetApp NICs that you're using, correct???
What type of switch gear are you using? There are documented issues and recommended "tweaks" to improve performance in Gb environments. I'd recommend you do a search on the NOW site on whatever kind of switch you have. Example: Search on CISCO brings up lots of info. Also, negotiation on NICs has been a problem in the past. May want to make sure it is "hard set" and not "auto negotiate".
Other factors will affect performance: - Clustering essentially cuts your NVRAM in half. This will not effect read performance but, it will effect write performance. - RAID and Volume layouts, drive types, size, BCS vs. ZCS, etc... will all play a part in performance.
Hope that helps.
Todd
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com]On Behalf Of Art Hebert Sent: Wednesday, April 03, 2002 8:26 PM Cc: 'Toasters (E-mail)' Subject: Question on gig slot assignments and fiber card assignments
I have a netapp f760 clustered running 6.01R3 with the following slot assignments:
slot 9 nvram slot 8 cluster interconnect slot 7 redundant fiber loop slot 6 intel fiber gig card slot 5 intel 100mb card slot 4 primary fiber loop
One 7 x 18gb fiber shelf, one volume.
I can't seem to get more than 10MB througput a little worse than on the 100mb card. The cpu numbers drop but so far the gig card isn't a big win.
I've tested from a sun 420 with a syskonnect fiber gig card in the fast pci slot. I've played with about every combo of negotiation, etc. Tested with a cisco 6408 gig card in a cisco 6509 switch and also tested card to card with just a fiber cable.
Here is sysstat:
time dd if=/dev/zero of=./100m.1 count=100000 bs=1024
2% 0 0 0 0 0 32 16 0 0 38 2% 0 0 0 0 0 0 0 0 0 38 2% 0 0 0 0 0 0 0 0 0 38 2% 0 0 0 0 0 32 16 0 0 38 2% 0 0 0 0 0 0 0 0 0 38 88% 1620 0 0 13874 327 496 260 0 0 38 87% 1216 0 0 10429 246 1016 10288 0 0 38 84% 983 0 0 8490 199 741 14470 0 0 38 77% 955 0 0 8190 193 668 13712 0 0 38 73% 956 0 0 8198 193 1121 8725 0 0 38 73% 955 0 0 8190 193 1380 10912 0 0 38 77% 956 0 0 8198 193 1586 10282 0 0 38 81% 1123 0 0 9570 227 488 9646 0 0 38 83% 1133 0 0 9718 229 1049 9810 0 0 38 85% 1224 0 0 10499 247 384 9954 0 0 38 CPU NFS CIFS HTTP Net kB/s Disk kB/s Tape kB/s Cache in out read write read write age 75% 855 0 0 7331 173 1522 12376 0 0 38 74% 534 0 0 4571 108 4268 13680 0 0 38 7% 0 0 0 0 0 116 3844 0 0 38 2% 0 0 0 0 0 32 16 0 0 38 2% 0 0 0 0 0 0 0 0 0 38 2% 0 0 0 0 0 0 0 0 0 38 2% 0 0 0 0 0 32 16 0 0 38
Questions:
1) is it better to have the primary loop in slot 8 and move the cluster interconnect? 2) how is everybody elses netapps laid out? 3) do i need more disk to get better performance?
-
Art Hebert -
Todd Phillips