"QoS" can mean different things to different people, but I believe the essence of what you want to show is that, during times of network congestion (ie, queueing), the delivery of higher priority traffic takes precedence over the delivery of lower priority traffic. (Conversely, when there is no congestion, there is no differentiation in treatment of traffic). This requires several things of your simulation/demonstration:
- An explicit QoS policy that objectively prioritizes classes of traffic and quantitatively establishes relative treatment of classes
- A means of identifying the relative priority classes of packets
- A means of enforcing the QoS policy on the traffic classes
- A means of generating traffic
- A means of congesting the network to demonstrate that the policy is actually working.
The demonstration could be as simple as two traffic sources sending to a single receiver, each on separate interfaces on a single switch/router. As long as the ingress offered load from the two sources exceeds the egress transmission capacity to the receiver (even very briefly), you will have congestion. Your policy could prioritize one source over another, one traffic class over another from one of both sources, or any combination that you can think of. If the received traffic matches your policy, you have success.
Cisco's MQC (Modular QoS CLI) provides the means to encode the QoS classification and enforcement into class-maps and policy-maps. While I have not used Packet Tracer, I cannot comment on its support of class-maps and policy-maps, but I can say that a ping alone is unlikely to produce a successful demonstration. That is, you are unlikely to see any difference in ping times with QoS applied versus without QoS applied in the absence of congestion. So you will need to ensure that there is adequate "background" or "noise" traffic present during your ping test so that there will be queueing taking place (without queueing, QoS queue management and scheduling are essentially null functions).
In real life, operators test the QoS capabilities of network elements by connecting them to special-purpose traffic generators (Spirent, Ixia, etc) and blast away. They then compare the received traffic to the configured policy to see if the QoS actually works as designed. A software-only demonstration can be used, as long as you are only looking for coarse results, not fine-grained.
Disclaimer: I am long in CSCO
