Storage VMotion Causes 100% CPU utilization on Filer

[jgunnarson]

We have been having a Filer issue whenever we do any sort of large data copy on our filer.  We have a FAS2050.  We have noticed this during Exchange Snapshots and whenever we migrate a VM's storage to another datastore.   We see 100% CPU utilization on the Filer as soon as we start the migration.  The base average on the CPU prior to the migration is 25-30%.  The datastores are iSCSI LUNS and not NFS.  It doesn't matter how much data is copied it happens for the smallest amount.  You definitely can notice a difference in performance on other VM's during this migration.

We do have a ticket open but they haven't found anything yet.  Initially they said our filer is overloaded but how can it be overloaded if the CPU is only using 25-30%?  We never had this problem with an HP EVA.

[radek_kubka]

Hi and welcome to the Community!

Storage VMotion is quite intensive sequential read operation, which in turn requires a lot of CPU power on NetApp. WAFL is fragmented almost 'by definition', which is fine for random workloads. For sequential ones, a complex read-ahead caching algorithm is used.

You can try improving things by checking & possibly reducing fragmentation via reallocate command.  Also, if your ONTAP version is lower than 7.3, I recommend upgrading it, as 7.3.x is more efficient with sequential reads.

Finally, FAS2050 CPU is really a slow one, I am afraid:

http://communities.netapp.com/message/62442#62442

Regards,

Radek