Network and Storage Protocols

FC vs LACP (1gigabit members) performance

erpadmin
2,365 Views

Short background info:  I am currently running Oracle database 12c on Solaris 11.2 SPARC with ZFS filesystem.  I am researching the idea of switching over to dNFS so I can leverage the many features that Netapp/SnapManager will afford.

 
I've thought about, at least for the short term, dedicating (2) 1 gigabit ports for LACP mode on both the FAS2552 and the resepctive Oracle T5 server involved.  Will dNFS and other operations be able to effectively use this LACP setup and be comparible to an FC setup?  Our current FC backbone is 4gb but we don't come close to even saturating 2gb (probably 70% at best).  One concern I have had is that the general throughput cap for a single 1gigabit interface is rougly 90megabytes per second and this netapp can get 120 megabytes per second write speed pretty easy.  Not sure about read speed -- haven't benched it.
 

Long term, i'm looking at adding 2 additional 10gbe interfaces on our FAS and maybe direct attaching the production T server or getting the addon modules for our 3750x stack.

Anyway any feedback on this is appreciated.

1 ACCEPTED SOLUTION

aborzenkov
2,336 Views
With LAN aggregate you depend on switch for load balancing. Cisco 3750 can distribute packets basing on MAC or IP, which means traffic between server and NetApp will always use single physical interface only. With dNFS you are better off to configure each physical link individually and relying on dNFS for load balancing and failover.

View solution in original post

1 REPLY 1

aborzenkov
2,337 Views
With LAN aggregate you depend on switch for load balancing. Cisco 3750 can distribute packets basing on MAC or IP, which means traffic between server and NetApp will always use single physical interface only. With dNFS you are better off to configure each physical link individually and relying on dNFS for load balancing and failover.
Public