MetroCluster Fabric - How the storage fabric (backend) works ?

ple · ‎2012-07-22

Hello,

How does the 2 fabric FC storage (2 x switchs Brocade 300E/site) works ?

I know the NVRAM traffic is carried on only one fabric (and in case of trouble, failover on the second fabric)

How the SyncMirror trafic is carried on ? is it carried on only one or on the two ? how it works ?

If on the two fabric: the traffic is it load balanced on the two or first carried on the first fabric and then on the second (when the first have lot of traffic), or other ?

Thanks for you help

Philippe

pascalduk · ‎2012-07-22

Both ISLs are used actively and Data ONTAP load balances the disk traffic just like with multi path on standard HA pair or single controller.

BTW, Data ONTAP actually load balances the disks on the 4 FC links it has to the 2 local site switches. You can actually see the amount of traffic on each link with the hostadapater stats command (ONTAP 8.x and diag mode). Short example output from a metrocluster (pool 1 is also used for reads):

Instance instance_nam node_name instance_uui total_reads total_writes bytes_read bytes_writte max_link_dat

/s /s /s /s KB/s

0c 0c 71 2 2680439 12455 1000000

0a 0a 75 2 3005309 12455 1000000

1a 1a 82 2 2783193 12455 1000000

1c 1c 66 2 3000638 12455 1000000

0c 0c 92 0 4236440 0 1000000

0a 0a 70 0 2533960 0 1000000

1a 1a 90 0 3530800 0 1000000

1c 1c 67 0 2326480 0 1000000

0c 0c 369 84 14839760 7163000 1000000

0a 0a 301 82 14385280 7096960 1000000

1a 1a 303 77 12389000 6098040 1000000

1c 1c 315 88 13357760 7228000 1000000

0c 0c 95 85 6013506 10071688 1000000

0a 0a 70 92 3822857 10309090 1000000

1a 1a 103 99 6781298 11525194 1000000

1c 1c 82 88 5537662 10354805 1000000