QOS on Cisco 1841 between MS TMG and managed Cisco 1841?

Question

Replicating our VM data from our Site A to a Hosted Provider (Site B) for DR purposes.
Crude annotaion of our network:
VIRTUAL HOSTS-----NORTEL L3 SWITCH-----MS TMG 2010 EDGE FIREWALL-----ISP MANAGED CISCO 1841-------------------CLOUD---------------SITE B
At times the replication traffic is hogging the connection and causing degraded performance for VPN clients amongst other things.
TMG 2010 doesn't support QOS and we cannot make any changes to the ISP managed Cisco router, nor can we request changes to be made.
My plan was to get a hold of a small switch that supported QOS and place this between the MS TMG and the managed router but we actually have another Cisco 1841 sitting doing nothing, would I be able to use the spare 1841 for this purpose?
Many thanks
Steve

Answer 1

Disclaimer
The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.
Liability Disclaimer
In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.
Posting
10 Mbps is around the limits of an 1841.  It might or might not have enough capacity.  You could try using it and monitor its CPU when it's dealing with saturated 10 Mbps.
You can police or shape before the 10 Mbps bottleneck.  That will insure you can guarantee bandwidth for the non-replication traffic.
You can only police after the 10 Mbps bottleneck.  If the replication traffic is TCP based, dropped packets should slow the sender, but the sender can often still burst saturate the link before it detects the drops and slows.  If you set a very low bandwidth allowance for the replication traffic, you can often keep it from burst saturating the link.  The other issue with policing, it cannot dynamically allow bandwidth utilization (i.e. prioritization) as you can do with a shaper.
If most of the replication traffic bandwidth consumption is one way, insert the 1841 anywhere upstream (where all traffic will pass through it) of the 10 Mbps bottleneck, and shape or police.  I would recommend shaping with a very low bandwidth allowance for replication (e.g. 1%).  This will allow replication traffic to use all 10 Mbps, but any other traffic will get priority.
If the replication traffic bandwidth consumption is two way, you'll really want a 2nd device, on the other side of the bottle neck.
PS:
BTW, in lieu of shaping, having your upstream (of bottleneck) also with a physical 10 Mbps interface works even better.  Again, de-prioritize the replication traffic.
e.g.
class-map replication
match
policy-map phy-10m
class replication
bandwidth percent 1
(optionally random-detect)
class class-default
fair-queue
int 10m
service-policy output replication