Introduction:TCP has responsibility of providing end-to-end reliable data on non-reliable Internet.This project is based on one powerful performance measurement technique called “Passive measurement”.
Passive measurement refers to the processing of measuring network, without infusing any traffic on network.Passive measurements used for monitoring traffic volume but can be used for network performance In contrast, some probe packets are introduced in Active measurement.
Active measurement, traffic injected through external sources for end-to-end data transmission, these artificially introduced packets may affect the performance of other applications between network paths..“IP networks supported only a best effort service”. TCP Operation and Performance:TCP was specifically designed to provide a reliable endto-end byte stream over unreliable internetwork and responsible for internet stability.TCP using basic technique as “positive acknowledgement with retransmission.
Reliability is basic issue in TCP that is achieved through a strong concept “sliding window”. The major problem with simple positive acknowledgement technique is, excessive amount of network bandwidth consumption.
RWIN is very important to calculate “Throughput” of network.Sliding window protocol solves the bandwidth problem by transmitting packets before acknowledgement received.
Sliding Window Mechanism and Window Size:The Sliding window performance can be judged through the size of window and the network speed on which packets are accepted. Intelligently selected window size can keep the network saturated all the time and idle time of network could be reduced to Zero, and obtains higher throughput than a simple positive acknowledgment protocol. The TCP window size is set to 64K initially because of its 16 bit header but it can exceed by using the window scaling factor options at the start of connection.TCP sender need to wait for at least one Round Trip Time (RTT) before acknowledgement received for transmitted segment. THROUGHPUT= (w. MSS)/RTTProfessor Jonathan Pitts quotted i.e Mathematically it can be presented like this, Max Throughput= Max advertised window/RTT.
RWIN is receiving window and equals to Bandwidth-Delay product. BandwidthDelay product can be calculated as Maximum available bandwidth multiplied by total latency.Congestion Control and Congestion Avoidance:Researchers agree that most of packets loss is due to congestion. It is important to know how TCP handles Congestion; a suitable window size is chosen.
SLOW START: Slow start algorithm removes the issue where two end point of communication transmit the data as fast as possible up to agreed window size completely disregard of network congestion.
CONGESTION AVOIDANCE :TCP uses this algorithm when packet loss occurs. Most of packet lost are due to Network congestion and shows there is high timeout and duplicate acknowledgements.
FAST RETRANSMIT:In Fast retransmit data continuously transmitted to keep throughput maximum and during this time it waits for acknowledgement of retransmitted packets.This mechanism can be monitored with any packet capturing tool. In this project it can be visualised and analysed in “wireshark”.
FAST RECOVERY:Principle behind fast recovery algorithm is to act as an interrupt; it stops TCP to go into slow start mode when it goes into congestion avoidance mode which is caused by Fast Transmit rather than simple timeout.Test Bed Description:The test bed used for conducting the experiments required assessing how TCP/IP behaves in different testing conditions.
TOPOLOGY DESIGN AND BASIC CONFIGURATION:There are two ways to put host in different network either by putting them in different domains Networks (dependent upon their network addresses) or by creating Virtual LAN. PC SYSTEM‟S SPECIFICATIONS:
PC1:PC1 is working as Traffic Source in most of the experiments. This System lies in VLAN4 with IP Address 192.168.4.2 .
PC2:PC2 is working as Traffic Sink in experiments and system is in VLAN5 and IP address is 192.168.5.2.
PC3:PC3 is very important system to check network performance in different extensive conditions and is a soul of project.System has two network interfaces for VLAN 4 and IP address is 192.168.4.1.
PC4:PC4 although doesn‟t work as main stream system in experimentation.
CISCO ROUTER 3640:Cisco 3640 router is used as multi service machine which is used from small to corporate sector and some internet service providers also used them.
IPERF is bi-directional traffic generation benchmark, although it is widely used for end-to-end performance measurement but it is not an official standard as it is open source. JPERF allows changing the buffer size but it is highly recommended not to exceed more than 130KB because after this it really doesn‟t affect on the network performance and its just waste of resource.
WIRESHARK: Wireshark is just like “tcpdump [27]” but with GUI interface and enhanced utilities. It is most popular network analyser is available on almost all Operating systems like Linux, Solaris, Mac, Unix-like and Windows. It understands the structure of different network protocols. it can capture packets with 500 capture and display filters. It allow user to save (.pcap) packet files and generate lots of statistical combination for file.Wirehsark GUI displays a TCP captured packet in three window pane.
NISTNET:Nist Net is useful in many terms as an inexpensive PC can emulate environment, works as router where router box is costly to buy and run. It has ability to perform complex router like scenarios by introducing traffic congestion, sever and sensitive delays, packet loss, bandwidth limitation, packets duplication and packet rearrangement.
PING TOOL:Ping tool is to determine the Host availability. It sends theEcho request to host and wait for ICMP packets which were sent to check host availabilityand gives the Round Trip Time (milliseconds) and Packet Lost(number, percentage) in summary.
Experimental Methodology: Technique which is adopted in this project is measuring through congestion window at sender to calculate RTT and Throughput, considered as popular passive measurement technique.
NISTNET AS NETWORK IN BOX:Nist Net sets up minimum and maximum threshold for DRD to drop packets. Nist Net also implemented Explicit Congestion Notification (ECN) through DRD by associating it with queue length less than ECN minimum threshold. EXPERIMENTAL RESULT: Many Experiments done to validate network performance with altered Emulation settings, TCP window size and Buffer sizes. Every experiment is done at least 5 times but two best possible performance values are taken for throughput calculation purpose.
Project and Methodology Limitation: Passive measurement need full network access or in other words it need to be assumed this is tester/researcher‟s “own network”. As passive measurement does not inject any traffic so it gets relatively less samples and inaccurate packet pair [25].
General Conclusion: Better solution is to increase network performance. In this project objective was to investigate lightweight measurement techniques.