". . . my replication link gets saturation almost 24H, so for that we are planned to increse the bandwithmy replication link gets saturation almost 24H, so for that we are planned to increse the bandwith . . ."
Insufficient information, to say for sure, but often unless you increase bandwidth by lots and lots, you often have the same issues (just paying more for your WAN bandwidth).
"once we have increase our bandwidth, i want to apply the Qos . . ."
Unclear why you need to wait to increase bandwidth to use QoS. Often, QoS is much, much better for mitigation than increasing bandwidth. (NB: that's not to say increasing bandwidth cannot be of benefit, just, often, QoS, is much more effective in many cases, like the one you describe. Also, the two, QoS and additional bandwidth, are not mutually exclusive.)
"can i apply the Qos on GRE encrypted Tunnels,"
Usually yes, but you haven't described your platforms.
"can i apply this on our interface connected to the ISP dispate we use encrypted GRE tunnel"
Also usually yes, and sometimes you want QoS on both tunnel and physical interfaces. In most cases, having QoS on the tunnel interface is "better".
"please can someone guide me to improve this"
Need much more information, about your environment, to provide good suggestions.
I will say, your device only QoS, usually can only "optimize" more than a two node connection "sacrificing" WAN bandwidth. Sometimes, though, a topology like yours can be "optimized", but much depends on where your traffic actually flows (i.e. to/from).
In a multipoint topology (with more than two nodes), Cisco does have a latter variant of DMVPN which does dynamic bandwidth shaping, which can go a long way in avoiding "sacrificing" WAN bandwidth. Personally, I've never used that variant, so cannot comment on how it appears to work in the "real world". (Also, again without knowing particulars of your platforms, don't know if your device currently or might support that DMVPN variant.)
