دانلود رایگان مقاله TCP بر فرصت انتقال کمیاب در شبکه رادیویی شناختی

Abstract

Transmission control protocol (TCP) is the most popular transport layer protocol for applications that require reliable and ordered data delivery essentially. In this paper we consider the deployment of TCP to secondary users (SUs) in overlay cognitive radio networks (CRNs), and address its performance degradation; in CRNs, SU’s transmissions are frequently disrupted by the detection of primary user’s transmission, and which makes the SU experience consecutive retransmission-timeout and its exponential backoff. Subsequently, the TCP in SU does not proceed with the transmission even after the disruption is over or the SU hands over to other idle spectrum. To tackle this problem, we propose a cross-layer approach called TCP-Freeze-CR; lower layer protocols send the overlying TCP two different cross-layer signals, freeze on the detection of primary user’s transmission, and unfreeze after handing over to an idle spectrum. Moreover we consider a practical situation where either secondary transmitter (ST) or secondary receiver (SR) detects primary user’s transmission; therefore additional message exchanges are needed between ST and SR to retrieve and resynchronize to other idle spectrum, i.e., spectrum synchronization. This situation is more complex than the case where both ST and SR detect primary user’s transmission. Hereby, we develop a spectrum synchronization procedure coupled with TCP-Freeze-CR into a finite state machine. All of our proposals are implemented and evaluated on a real CRN consisting of 6 software radio platforms. In the implementation, we deploy 802.15.4 implementation as a target physical layer protocol, and couple it with TCP-Freeze-CR using Unix Domain Socket. The experimental results illustrate that standard TCP suffers from significant performance degradation in CRNs, and show that TCP-Freeze-CR can greatly alleviate the degradation; e.g., for 1200 s, ST with TCP-Freeze-CR can send about 10 times more packets than ST with standard TCP.

8. Concluding remarks and future work

In this paper, we address the problem that standard TCP in SU side encounters in overlay-CRNs; primary transmission disrupts SUs’ transmission, which results in drastic decrease of TCP performance following consecutive RTOs and exponential backoff of RT. Inspired by Freeze-TCP, we propose a cross-layer approach called TCP-Freeze-CR to tackle this problem. We verify the performance improvement by implementing our approach on a real testbed consisting of the software radio platforms, USRP E100. In this work, we consider only single-hop network. However, we need to show that our approach can be extended to multi-hop network. It is also possible to couple the cross-layer mechanism with one of TCPs for mobile ad-hoc network (MANET) or other TCP variants that consider fair bandwidth sharing among multiple connections (e.g., TFRC). We consider overlay-CRN only in this work. Our approach can be further developed for supporting underlay-CRNs coupled with transmission power control mechanism. Also it is possible to deploy proactive spectrum sensing that can relax the constraint that PUs are tolerant on the message exchanges for the spectrum synchronization, and incorporate packet loss differentiation and link layer retransmission schemes in order to achieve faster packet recovery right after ST is unfrozen. It is also worthwhile to study a TCP regulator scheme, whose requirement has already been addressed in [36], in order to cope against the high jitter inflicted by consecutive freeze and unfreeze operations. Last but not least, we have considered here a simple energy detection mechanism only for detecting primary transmission, and configure the experimental network where SUs never fail to detect primary transmission. We admit that this setup is far from reality, and it is more valuable to consider more complex and realistic CR environments with false alarming and detection fail, which definitely yields poorer TCP performance or unwanted interference to primary transmission. Therefore we need to consider imperfect spectrum sensing and couple more advanced spectrum sensing scheme such as statistical hypothesis testing [38] with our TCP.