Abstract — Transmission Control Protocol (TCP) is the core of the Internet carry out the communication and delivery of all types of data and connect millions of computers around the world. In this paper an investigation of the performance of TCP flavors namely Reno, NewReno, Vegas and SACK based on the Bandwidth performance and the actual throughput. The network topology is a satellite network and it had been configured into different scenarios to enlarge the chances to get the desired goal. Simulation methodology is used in this study using OPNET simulation tool. The objective was to investigate and find out the performance of TCP variants according to the bandwidth in satellite network. However, the results are shocking, TCP Reno is the maximum aggressive, and greatest amount of throughput. In TCP NewReno, it follows Reno's steps by becoming the second maximum aggressive, and second highest throughput. SACK is reasonable to Reno and NewReno, but when it is compared with Vegas, it clears that it is very unfair. Lastly, Vegas shows the highest degree of fairness (minimum aggressive) and also Vegas produces the lowest amount throughput.
Keywords—TCP/IP Protocol; TCP flavors; satellite networks; OPNET Simulation ; Throughput
Certainly, TCP is a component of TCP/IP internet protocol suite. However; it is definitely considered as an independent, general purpose protocol since that it can be used with other delivery systems. TCP protocol makes very minor assumptions about the underlying network, for example it is possible to employ TCP over single network just like an Ethernet or even over complex networks such as the global Internet . Indeed, it is the dominant transport protocol over both wired and wireless links .
The problems that arise when a Network underperforms are exceptionally huge. There is Internet underperforms in the public and private organizations across the whole industries that make use of Internet. This will be affected in terms of the cost, the quantity and the quality of services offered.
Today TCP/IP networks are used frequently and many people experience it from fixed locations (home, school, and office) or from mobile devices while people are on the move . Poor performance of TCP/IP networks will absolutely not be tolerated.
Since there are many different flavors (variants) of TCP, then definitely their performance differs from one flavor to another and from one environment to another. If the TCP engine is not optimized then the high performance of TCP/IP network will be greatly hampered. There are more than nine (9) concepts that can be associated with the performance of TCP and one of these concepts is Bandwidth or effective throughput.
Bandwidth or effective throughput is one of the metrics of the TCP performance. It means the number of applications bytes sent over the network or transferred in seconds and when it comes to the transfer of large files then the effective throughput application (speed or bandwidth of the connection) becomes a key performance measure.
In this paper the comparison of Reno, Vegas and NewReno TCP flavors is carried out in a satellite network topology which changes its network architecture into six different configurations and study the effective throughput application of each TCP flavor.
II. Current status of TCP
The latest TCP related research was the RFC 8257 "TCP Congestion Control for Data Centers " dated on October 2017. Generally Quick-Start identifies discretionary mechanism not only for TCP but also for other transport protocols, in cooperation with routers, to resolve a permitted sending rate at the start and, sometimes in the middle of data transfer (e.g., following an inactive period of transmission). At the same time as Quick-Start was intended to be used by a range of transport protocols, in RFC 4782 it is only specific for the TCP. Quick-Start is designed to permit connections to make use of higher sending rates if there is significant idle bandwidth along the path and provide that the sender and all of the routers along the path approve the Quick-Start Request.
III. TCP Variants
The following is a list that shows some of the TCP flavors or variants:
- TCP Reno and TCP NewReno
- Selective Acknowledgement (SACK)
- Forward Acknowledgement (FACK)
- TCP Vegas
- TCP Veno
- TCP Tahoe
In this paper four TCP flavors have been chosen, and they are as follows:
A. TCP Reno, it is classical just like its predecessor TCP Tahoe and that makes it familiar and widely used. TCP Reno brought primary enhancement over Tahoe; because of that Reno incorporates a new mechanism that Tahoe does not support. This new mechanism is known as the fast-forward algorithm.
B. TCP NewReno, this is one of the most recent flavors and by definition it a TCP resultant variant from the modification of Reno. Its most important feature is that it improves retransmission process during the fast recovery phase of TCP Reno. NewReno can detect the occurrence of multiple packet loses [3, 4].
C. TCP Vegas, in 1994 this flavor has been presented after Reno and before NewReno. Vegas does not wait for loss of trigger congestion window reductions and that makes its fundamental difference from other TCP flavors .
D. Selective Acknowledgement (SACK), this TCP flavor is an approach that corrects the problematic behavior of multiple dropped segments that reduces the overall throughput. With selective acknowledgments, the data receiver can notify the sender about all segments that have arrived successfully, therefore allowing the sender to only retransmit the segments that have really been lost .
IV. OPNET Simulation tool
OPNET stands for Optimized Network Engineering Tools, and was created by OPNET Technologies, Inc., which was founded in 1986. OPNET is a network simulation tool set. It is powerful and can create and test large network environments via software .
OPNET Modeler is a very useful tool for network modeling and simulation. The Network simulation provides a way to model the network behaviors by calculating the interactions between modeling devices. Discrete event simulation (DES) is the typical method in large-scale simulation studies instead of a simpler time-based method. OPNET is capable of simulating in both explicit DES and hybrid simulation modes, and supports other simulation features like co-simulation, parallel simulation, high-level architecture, and system-in-the-loop interactive simulations.
V. Simulation Topologies
In this paper the simulation of the following network topologies or configurations will be carried out as shown in figure (1). At the beginning a simple network topology of two combined TCP variants will be simulated and studied. The combination of the four TCPs can be found by the following variations:
- Reno versus Vegas
- Reno versus NewReno
- Reno versus SACK
- Vegas versus NewReno
- Vegas versus SACK
- NewReno versus SACK
This is the start of the simulation topologies and as shown in Figure (1) a network of four nodes, two flavors of TCP Reno and two TCP Vegas is devised each side of the two branches of the network. This is the first scenario of the simulation topologies. There exists receiver side TCPs known as TCP Sinks which correspond to the sender side TCPs. In this Topology and every other Topology that will be simulated.
Figure 1. Simulation Topology for TCP flavors
Other scenarios will follow the same configuration of nodes and parameters as in table 1 and so on and so forth until the sixth scenario is reached-the combination of One NewReno and one SACK1. The same topology is repeated and all the other with the fixed parameters such as the bandwidth at router and as well as at the Nodes.