Improving Throughput in Lossy Wired/Wireless Networks
- PDF / 1,231,023 Bytes
- 12 Pages / 439.37 x 666.142 pts Page_size
- 37 Downloads / 234 Views
Improving Throughput in Lossy Wired/Wireless Networks Taha Saedi1 · Hosam El‑Ocla1
© Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Wireless communication is more prone to random loss than wired communication because of noise and mobility. Over years researchers have developed TCP variants that do not decrease the send window when random loss arises.Years ago it was introduced TCP CERL algorithm that proved to present a high performance compared to other protocols. Here, we test CERL assuming two-way transmission of relatively heavy load and compare with TCP BIC, TCP NewReno, TCP Westwood+, TCP NewJersey and TCP Illinois. Simulation Results show that TCP CERL gains a 145%, 137%, 120%, 97% and 125% throughput improvement over New Reno, Bic, Westwood+, New Jersey and Illinois, respectively. Keywords TCP · Random · Congestion · Wireless · Networks · Loss · Piggybacking · NS2
1 Introduction Reno algorithm is a descendent of TCP Tahoe to control congestion over computer networks. TCP Reno proposed a fast recovery mechanism [1], which means that if the TCP Reno sender receives the duplicate ACKs, TCP Reno enters to a fast recovery and remains there as long as duplicate ACKs continue to arrive rather than back to slow start that is in TCP Tahoe [2]. One of the main drawbacks of TCP Reno is its inability to recognize the reason of packet loss if it is owing to congestion loss or random loss and considers all loses are produced due to the congestion. Several attempts were presented not to decrease the send window if the random loss is detected. In [3], authors have assumed a receiver that can recognize the cause of the loss through using the gaps between arrival times. This solution postulates having only single wireless portion in the network topology while selective acknowledgment (SACK) was used to block the numerous losses. Excessive time needed to process SACK algorithm downgrades the performance particularly when the number of users increases. In [4, 5], Explicit Bad State Notification (EBSN) was developed. EBSN is mainly aim to update the TCP timer at the TCP sender to prevent decreasing the congestion window. This means that if a packet is failed to be sent correctly over the wireless link from * Hosam El‑Ocla [email protected] Taha Saedi [email protected] 1
Computer Science Department, University of Lakehead, Thunder Bay, Canada
13
Vol.:(0123456789)
T. Saedi, H. El‑Ocla
the base station because of congestion, the base station sends EBSN to the original sender to reset its timer to its previous state. There is an obvious improvement in the throughput using EBSN mechanism in the wireless links but yet not that adequate. In [6], TCP NewReno was proposed. The main objective of TCP NewReno is to examine if multiple packets have been lost in the congestion window once duplicate ACKs are received. In this case, lost packet will be resent until non-duplicate ACK is received. Others recommended the use “Binary Increase Congestion” TCP to improve congestion window performance b
Data Loading...
