1. An IP-based multimedia traffic generator
Faculdade de Tecnologia - UNICAMP
An IP-based multimedia traffic generator
Fernando L. Pinotti *
Tito R. B. Oliveira
Edson L. Ursini
Varese S. Timóteo
* Supported by CAPES

2. Outline - Introduction - Traffic Model - Distribution Function
- Architecture
- Simulation and Emulation
- Final Remarks

3. Introduction
- In this work we present an real time IP based multimedia traffic generator that creates packets with given patterns based on well known traffic distributions.
- A multimedia service is obtained by creating two or more instances providing different traffic profiles.
- In our approach, the multimedia traffic is emulated instead of simulated.
- Results are compared with the ARENA simulator software for validation.
In this work we present an real time IP based multimedia traffic generator that creates packets with given patterns based on well known traffic distributions.
- A multimedia service is obtained by creating two or more instances providing different traffic profiles. By using stream and elastic traffic patterns.
- The Multimedia Traffic Generator results are compared with the ARENA simulation software to ensure the efficiency .
- It is important to note that, in our approach, the multimedia traffic is emulated instead of simulated, since we generate, and send over the network, real IP packets. We use the IP headers, type of service (IPv4), traffic class and flow (IPv6) to distinguish the information type inside the packet. Which allows a multimedia gateway to separate the traffic.

4. Traffic Model
- Elastic Traffic: Accounts for non-interactive applications, like file transfer, , internet browsing, where the critical point is the integrity of the information, and delays don’t compromise the services and applications. For this propose we use the TCP protocol.
Data
FTP
SMTP, Etc.
- Stream Traffic: Accounts for interactive applications that require real-time delivery. The stream denomination is due to the characteristic of data packet flow that are nearly constant during a given traffic period. For this propose we use the UPD protocol.
Video
Voice
Audio, Etc.
- Elastic Traffic: accounts for non-interactive applications, like file transfer, , internet browsing, where the critical point is the integrity of the information, and delays don’t compromise the services and applications. In this case, the transmission control protocol (TCP) is used to ensure the delivery of the data and to perform the retransmission of segments in case of packet loss. Some extra delay is added in exchange of the guarantee that the packets will be delivered.
- Stream Traffic: accounts for interactive applications that require real-time delivery. The stream denomination is due to the characteristic of data packet flow that are nearly constant during a given traffic period. It is also known as inelastic traffic, due to the fact that it doesn’t support the delay variations as the elastic traffic does. To reduce the package transmission interval time, the User Datagram Protocol (UDP) is used, since there is no connection established between sender and receiver. Therefore, this protocol doesn’t guarantee the delivery of the packets, but this kind of service is more sensitive to packet loss than delay. The real time transfer protocol (RTP) can then be used to enhance the packet transmission.

5. Distribution Functions
- Exponential –Most used to analyze the interval between events.
- Log-Normal – Most used for a time estimative of fault repair.

6. Architecture Service Selection: New services are initiated.
Traffic Profiling: Provides information about the type
of traffic that will be generated for each service.
Packet Generation: The final state of the state machine,
where the packets are built and sent.

7. Architecture Packet Building
The TCP packet size depends on the amount of available bandwidth. For UDP, the segment is filled with information and its size depends on the codec rate.
In order to obtain some preliminary multimedia traffic, we considered the IP packet to be created through a socket.
Once the system is calibrated for the basic services, the socket interface will be replaced by a full TCP/IP stack integrated in our system. Then we will be able to use the fields of the IP header, in particular, the Type of Service (Ipv4), Traffic Class (IPv6) and Flow Label (IPv6) to separate traffic by service (class) and by logic channel (flow).

8. Simulation and Emulation
- For the propose of testing the traffic profiling module of the traffic generator, the transmission was made using Loop-back in Scenario 1 and Ethernet in Scenario 2.
- The inter-arrival service process and its duration were estimated using the exponential distribution function, but other types of distributions could be used.
- In order to validate our emulations, we used the ARENA simulator for comparing the results.
- For the propose of testing the traffic profiling module of the traffic generator, the transmission was made using Loop-back in Scenario 1 and Ethernet in Scenario 2.
- The inter-service arrival process and its duration and size were estimated using the exponential distribution function, but other types of distributions could be used. according to the network behavior that one wants to emulate. We can use simulations to perform a preliminary study of the network behavior, and translate it into a distribution function that better represent the packet flow and services of the network we want to emulate. The distributions are then selected in the emulator that will generate the packets accordingly.
- In order to validate our emulations, we used the ARENA simulator for comparing the results. Arena was developed for discrete event simulations and consists of modular blocks where we are able to specify distribution parameters and simulate the same service environment that we set in the traffic generator.

9. Simulation and Emulation
Scenario 1
-Bandwidth fixed at 384 kbps.
- Loop-Back.
- 10 hours of emulation.
- Exponential distribution.
Comparison between ARENA simulator and Traffic Generator results.
Number of generated services.

10. Simulation and Emulation
Scenario 2
- Bandwidth fixed at 4000 kbps.
- Ethernet
- 10 hours
- Exponential distribution
Comparison between ARENA simulator and Traffic Generator results.
Number of generated services.

11. Simulation - Scenario 3
-Bandwidth fixed at 384kbps.
At scenario 3, the Arena simulations were performed for video services, in order to observe their behavior as the traffic variables change.

12. Simulation – Scenario 3
When the mean inter-arrival time increases, the number of both completed and the rejected services decreases as expected. The mean service duration is set to 900 seconds.
The figure shows the number of total, completed and
rejected services as a function of the mean service inter-arrival time for a fixed mean service duration and maximum number of simultaneous services. When the mean inter-arrival time increases, the number of both completed and the rejected services decreases as expected

13. Final Remarks Thanks you!
- The next version of the multimedia traffic generator will contain an integrated TCP/IP stack. In order to:
Remove the time spent by the socket interface.
To allow the manipulation of the IP header.
Thanks you!