[c-nsp] Application Protocol Performance in low latency envrionments

[Ash Net]

Hi Folks,

Sorry about the OT here, I'm looking to get some feedback regarding
some of the most common application protocols (CIFS, NFSv3,SQL net,
Snapmirror, ndmcopy) used in most Enterprise envrionments  and their
behavior in a sub msec campus latency environments vs ~3 msec latency
over 10G Lanphy environments in metro DC topologies and potential ways
to speed them up.

In our lab testing using wan simulator and injecting 3msec latency
(delay between 2 sites), we've noticed that all the above protocols
take a considerable performance hit with when even minimal latency is
introduced. The file/data transfers/transactions take almost twice as
long or sometimes even more and the protocols aren't able to ramp up
to use all the bandwidth available (Gig terminations on the src/dest
boxes sitting idle). This behavior is completely different than what
was noticed by removing 3 msec latency and running off a sub-msec
configuration where most of the above protocols were able to ramp up
and utilize a fair share of b/w available to help achieve faster
file/data transfer rates/transaction times which is to be expected.

Question is, is this something that other environments normally
experience as well with the above listed protocol set where they are
lightning fast in a sub-msec topology and slow in 3-5 msec
environments, and if they do, what are some of the techniques that are
being used today in the Industry to optimize these protocols and help
achieve lan like performance over a low latency (3-5 msec) metro
ethernet dwdm based WAN.

WAN ACCeleration is normally a solution to speed up transfer rates but
3 msec isn't a big enough delay to perhaps introduce it, do other
environments deploy WAN Accl in such environments anyways

The reason for performance degradation solely seems to be latency
related since there's tons of b/w available in the lab setup and over
10G lanphy paths. Do people still  deploy QOS for better traffic
management on the lanphy interfaces even with no saturation?.

Overall I'm looking for some feedback and suggestions If possible as
to how the industry is currently able to better utilize and saturate
low latency based 10Glanphy circuits between metro DC locations and
speed up protocols by somehow allowing them to ramp up and use as much
of the b/w that's currently available.

Any feedback would be greatly appreciated and thanks in advance.