2010-12-09 11:03 AM
I have a High Availability FAS2020 dual heads. I am wanting to use this device as an ISCSI SAN that will house Several Virtual Machines for our Organization. The most resource intesive is a Database Server, and our Exchange Server.
I am a little disappointed that I need to Carve up so many disks just to run Data Ontap on both of the heads. My goal was to run just one large aggregate. But it appears that I need to divided the Disks between the two heads before setting them up in a Cluster.
What is the best way to do this and achieve performance? How many Disks get assigned to Each head. Raid4 or raid DP?
My second question has to do with the Networking ports. I need to be able to do ISCSI for my SAN. But I also need to Connect to my LAN for CIFS. It is imperitive that my solution be completely redundant, and still have decent performance. If I dedicate two of my 4 ethernet ports just to CIFS then I feel that the ISCSI SAN will suffer greatly from the lost bandwidth. Is there a way to use VIFS, with VLANS to get throughput and still get the end result that I am looking for?
THanks in advance.
Solved! SEE THE SOLUTION
2010-12-09 12:22 PM
Well the "HA" means something and costs something.
I have the same basic setup you are going for.
We use one nic for wan and one for iscsi and we set the nfo option on the nics to force failover if either nic fails.
If one GigE nic isn't enough then the FAS2020 is the wrong hardware.
You can push a lot of data over a dedicated GigE port.
Look carefully at block sizes and determine if jumbo frames will help or hurt.
I did a FAS2050 FC connected vSphere server last year too.
Because of a switch compatibility issue we went live on iSCSI and converted to FC after the firmware upgrade.
We did testing banging vm's against iSCSI and FC and our test results were inline with the NetApp 7% benchmark differences.
Best Practices for vSphere says NFS over iSCSI. I haven't done that yet but TR-3839 TR3749 and TR-3808 are recommended readings.
As for disk configurations:
If all you want is a backup head for active/passive then 5 data, 2 parity, 1 spare on one head and 1 data 2 parity and 1 spare on the other.
Yes, you can do raid4 on the passive head and get to 6 data on the active head. I don't recommend it but I've done it.
For active/active 3+2+1, 3+2+1 means 50% disk overhad. Add in WAFL overhead and you are at 45% of sales capacity.
Back to the first sentence. HA costs. If you don't need it 12 drives on one filer is a lot more spindles than 6.
Active/Active gives you twice as many active nics and cpu's but half the spindles on each filer.
2010-12-09 04:35 PM
I'm a big fan of using RAID4 with smaller filers for greater utilization.
I hope this response has been helpful to you.
At your service,
(P.S. I appreciate points for helpful or correct answers.)
2010-12-10 02:31 AM - last edited on 2015-09-10 04:29 PM by allison
Hi and welcome to the Communities!
I am a little disappointed that I need to Carve up so many disks just to run Data Ontap on both of the heads. My goal was to run just one large aggregate.
Have a look at this good post from Adrian how to maximize the capacity in your case:
I need to be able to do ISCSI for my SAN. But I also need to Connect to my LAN for CIFS. It is imperitive that my solution be completely redundant, and still have decent performance.
In a case of 2020, due to only 2 IP ports per head, you cannot separate IP SAN from CIFS on a physical basis, whilst having full redundancy. If you have a couple of switches supporting cross-stack LACP (like Cisco Etherchannel), you should create a multi-mode VIF on each head spanning both ports & then separate iSCSI from CIFS using VLANs.
2011-12-02 02:28 PM
I wasting a lot of space in a 2020
9 Disk Controller 1 - DP - 3.4 Space to use
3 Disk Controller 2 - RAID 4 - 600 Space to use
Is it OK to use to reduce vol0 on both controller where Data ONTAP resides let's say to 100 GB to maximize space?
also Is it OK according to your recommendation to convert my DP Raid in Controller 1 to RAID4....?
thanks a lot