[c-nsp] WAAS and minimum latency

[Eric Girard]

Tim,
        While in theory you should still see some improvement from CIFS with a setup like this, I've done a PoC/trial with a near identical setup, 1G/3-4ms latency, and the performance improvements where minimal at best.  The one caveat was the CIFS shares were being used by a questionable financial application and the average filesize was small, but in the end, the price/performance was impossible to justify given the size of WAE needed to handle that much traffic.
        In the more 'traditional' WAAS space above ~20ms of latency I've had great results every time.

Eric Girard

-----Original Message-----
From: cisco-nsp-bounces at puck.nether.net [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of James Michael Keller
Sent: Wednesday, July 15, 2009 10:41 AM
To: Tim Durack
Cc: cisco-nsp at puck.nether.net
Subject: Re: [c-nsp] WAAS and minimum latency

Tim,

I doubt you will see improvement over 3ms for general latency reduction
(assuming a OCX P-t-P link?).  However it will improve CIFS performance
if the files are being accessed and changed a lot by the users at the
site remote from the CIFS server.   The WAE on the server side of the
link will cache operations locally.   So say you move a file between
CIFS shares, normally that comes back through the client and back down
to another share.    With the WAE unit it will proxy that operation and
the operation completes at local LAN speed instead of WAN speed through
the remote client and back to the other server.   While WAE's will
fiddle with TCP settings to improve some performance, the main function
in the current release code is the data reduction features.  Either the
raw DRE caches or application level proxies (CIFS/MAPI,NFS, etc).
Latency may not improve, but effective speed and bandwidth will go up.

For our MPLS connected sites in the 50ms+ range, there is some
improvement of the RTT of around 40% on average across all the sites.
Traffic reduction runs an average of 30% with Content and version
management protocols and CIFS/MAPI making up the bulk of the traffic
reduction (all above 50%) .  The main non-optimized traffic is internet
bound in our case, as we centrally route internet out a data center from
the MPLS connected sites.

---
James Michael Keller



Tim Durack wrote:
> Anyone got figures on the *minimum* latency the various WAN accelerators can
> improve on?
>
> I ask as I have a customer with a couple of sites connected via GigE. RTT
> for SiteA -> SiteB is around 3ms. Migrating services between sites has
> reduced performance for some users (appears that SMB/CIFS is most affected.)
>
> I'm looking to see if I can "fix" things with WAAS, just not sure they are
> really designed for this scenario (I'm not a fan of WAAS, but if it fixes a
> problem...)
>
> Thanks,
>
> Tim:>
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>
_______________________________________________
cisco-nsp mailing list  cisco-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/