Stability and Performance Analysis of TCP variants using OMNET-Computer Sciences and Applications-Project Presentation, Slides of Applications of Computer Sciences

Download Stability and Performance Analysis of TCP variants using OMNET-Computer Sciences and Applications-Project Presentation and more Slides Applications of Computer Sciences in PDF only on Docsity! 1 Road Map  About The Project  Previous Work  TCP  TCP Congestion  TCP Congestion Control Mechanism  TCP Variants  TCP Tahoe  TCP Reno  TCP NewReno  TCP Vegas  Comparative Analysis of TCP Vegas with TCP Tahoe etc Work Plan  Questions docsity.com 2 About The Project  Title  Stability and Performance Analysis of TCP variants using OMNET++  Nature of the Project  A research and development project  Research and Development Work  Development of Simulation of Selected TCP Variant  Performance and stability analysis of selected TCP Variant  Results  Respective Study Areas  TCP  TCP Congestion Control Mechanism  TCP variants  Learning of OMNET++  Tools  OMNET++  Any other tool such as Network Simulator (NS) if needed docsity.com 5 TCP Congestion  Why Congestion Occur?  Reason 1 • Data arrives on a a fast LAN and gets sent on a slower WAN  Reason 2 • Multiple input streams arrive at a router whose output capacity is less than the sum of the inputs  Reason 3 • Load of incoming packets increased so that network couldn‟t handle the traffic of packets docsity.com 6 TCP Congestion Control Mechanism  TCP Congestion Control Mechanism  Prevents a source from exceeding network capacity by allowing to adapt certain transmission rate  There are four basic algorithm for TCP Congestion Control Mechanism.  Slow Start  Congestion avoidance  Fast Retransmit  Fast Recovery docsity.com 7 TCP Variants  Depending on these algorithm there are different TCP Variants  TCP Tahoe  TCP Reno  TCP NewReno  TCP Vegas etc docsity.com 10 TCP Tahoe Cont…  Congestion Avoidance in Tahoe  Tahoe uses “additive Increase Multiplicative Decrease”  In case of congestion, saves half of the window as ssthresh value  Then set CW to 1 and starts Slow start until reaches the ssthresh  Then increments linearly until encounters a Packet loss  Increases window slowly as it approaches the bandwidth capacity docsity.com 11 TCP Tahoe Cont…  Problems in TCP Tahoe  It takes a complete long timeout interval to detect a packet loss  Follow „Go back n‟ approach because it sends cumulative ACK‟s not immediate • Every time a packet is lost it waits for a timeout and the pipe line is emptied  This is major cost docsity.com 12 TCP Reno Retains the basic principle of Tahoe  Slow start and Re-transmit Timer Intelligent because  Lost packets are detected early  Pipeline is not emptied every time a packet is lost Improvement over Tahoe by detecting loss Via retransmit timer docsity.com 15 TCP Reno Cont…  Problems in TCP Reno  In case of multiple packet losses  We reduced window size twice for packets lost in one window  If window size is very small, when the loss occurs then we would never receive enough duplicate ACK‟s for a fast retransmit and we would have to wait for timeout. So couldn‟t detect multiple losses docsity.com 16 TCP NewReno Modification of Reno  Detect multiple losses Also Enters into Fast Retransmit Differ from Reno as it doesn‟t exit fast recovery until all the data is acknowledged.  Multiple retransmissions Overcome the problem of reducing the CWND multiple times docsity.com 17 TCP NewReno Cont… Notes the maximums segment (outstanding) after entering Fast Recovery When a Fresh ACK is received, two Cases If it ACK‟s all Segments that were outstanding, Exits and set CWND to ssthresh and continue Congestion avoidance If it is partial ACK, deduces that the next segment was lost, Retransmits that segment and sets the number of duplicate ACK‟s received to ZERO, exit after all the data is acknowledged docsity.com 20 TCP Vegas Cont…  How it is efficient  Suggest an Algo which checks for timeouts at a very efficient schedule  Doesn‟t require enough duplicate Ack‟s to detect packet loss  Detect congestion before the packet losses occur  Introduces Modified slow start which prevent it from congesting the network  Motivates three techniques to increase throughput and decrease losses docsity.com 21 TCP Vegas Cont…  New Retransmission Mechanism  Reads and Records the system clock each time a segment is sent  When an ACK receives • Reads the clock again • Does the RTT calculation • Uses this estimation decide to retransmit in the following two situations docsity.com 22 TCP Vegas Cont…  Two situation  When Duplicate ACK is received If (time – Current segment transmission time) > timeout value (RTT)  yes then retransmit without waiting for n(3) duplicate ACK’s • May losses are so great be OR the window so small (never receive 3 dup ACK)  When a Non-Duplicate ACK is received If it is first or second one after the retransmission, same condition,  if yes then retransmits the packet • This will catch any other segment without having to wait for a duplicate ACK  Still contains Reno Coarse grain timeout  Only decreases the CW if • The retransmitted segment was previously sent after the last decrease  Detect multiple packet losses docsity.com 25 TCP Vegas Cont…  Modified Slow Start  Reason of modification  If network bandwidth increases OR if we encounter burst of traffic from another connection  Incorporate congestion detection mechanism into slow start  It allows exponential growth only every other RTT  In between, CW stays fixed so valid rate comparison can be made  When actual rate falls below the expected by certain amount, we call this amount threshold,  Exits slow start and enters congestion avoidance mode docsity.com 26 TCP Vegas Cont…  So adding congestion detection to slow start is important. Vegas just offers beginning but still there is a problem  During slow start, TCP sends segments at twice the rate  If there aren‟t enough buffers, Vegas‟s slow start with congestion detection may lose segments before getting any feedback  For future, there are few solutions • Use rate control during slow start • Slow down as we reach available the bandwidth docsity.com 27 Performance Analysis of TCP Vegas with Tahoe, Reno & New Reno Over TCP Tahoe Over TCP Reno Over TCP New Reno Detect and Retransmit lost packet much sooner than Timeouts in Tahoe Prevents timeouts. It detects and retransmits more than one lost packet before timeout occurs Prevents timeouts Has fewer retransmissions as It doesn’t empty whole pipe whenever loses packets Doesn’t wait for 3 duplicate packets so it can retransmit sooner Doesn’t wait for 3 duplicate packets so it can retransmit sooner docsity.com