Qos for specific source

[Sipl_24034]

Hi,

 I have 2 Mbps internet connection at spoke

 And my Requirement is  (x.x.x.a /32 and x.x.x.b/32)
 my database servers which is on hub location ,
 when they will generate traffic  towards spoke they must get 80% bandwidth out of 2 MB

which i would like to configure only on Spoke router

How can i do that with QOS ?

[Georg Pauwen]

Hello,

I don't think this is possible. It wouldn't matter if the traffic would get policed or shaped at the spoke location, because it would have already traversed (and possibly fully saturated) the 2MB link.

What is the problem with configuring the QoS policy on the hub ?

[Sipl_24034]

As on Spoke we having database traffic and video  traffic in case video is going on at that time database unable to run Package and synk.   

And for video we having secondary preferences our main priority is Database

Let me know what can be possible solution ?

[Georg Pauwen]

So why not limit the bandwidth of video in case of congestion ? That could be implemented at the spoke location. What device do you have, can you post the config ?

[Sipl_24034]

That can be do but it must not be permanent policy .we can restrict video traffic on time of Network congestion scenario only

we having CISCO2921/K9  router

and in config we don't have anything specific

Two MPLS link with Static(No BGP) . one Lan.

[Sipl_24034]

As on Spoke we having database traffic and video  traffic in case video is going on at that time database unable to run Package and synk.   

And for video we having secondary preferences our main priority is Database

Let me know what can be possible solution ?

[Joseph W. Doherty]

You cannot effectively guarantee bandwidth, from hub to spoke, by only applying QoS at spoke.  What you want to do, is implement QoS at hub.  Is there a reason this cannot be done?

If you can implement QoS at the hub, and if the hub has more than 2 Mbps of bandwidth, you would shape all traffic (to that spoke) at 2 Mbps and then provide your database traffic 80% of the shaped bandwidth.

(BTW, I'm assuming the spoke will not receive ingress traffic other than from the one hub.)

Oh, I should mention, it's possible to police ingress traffic at spoke from hub, so that non-database traffic is limited to 20%.  The two major problems with this approach are:  first, generally, non-database traffic would be policed at 20% even when the database traffic doesn't need its 80%.  Second, transmission bursts can still congest the link before it hits the spoke's ingress policer.