Who says CIFS has to perform worse than NFS?
That’s demonstrably false and EMC’s current (And first) SPECsfs2008 benchmarks prove it here and here.
Look at the other benchmarks on the lists and compare the CIFS and NFS numbers. CIFS performance can be anywhere between two and a half to three times slower than the submitted NFS number for the same hardware/software configuration.
That’s assuming they even list a CIFS number and judging by how light the SPECsfs2008 CIFS list is you have got to believe there are more than a few CIFS results that were so embarrassingly poor they didn’t make it out of the data centre they were tested in.
Celerra on the other hand has proven that even above 100k IOPS there should be little difference between CIFS and NFS numbers if you’re doing it right.
A lot of vendors obviously aren’t doing it right.
There was no funky pre-release code or six months from delivery hardware at play here, just currently shipping code and hardware you could buy and set up in your data centre in 2009.
NAS users: CIFS isn’t a dog of a protocol.
The other guys just made a dogs dinner of implementing it.
