W. Richard Stevens. TCP/IP Illustrated: The Protocols, volume 1. Addison Wesley, 1994.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....information, there is no better source. There are many good texts that cover basic concepts of computer networking [3, 4, 11] The great technical writer W. Richard Stevens developed a whole series of classic texts on such topics as advanced Unix programming [5] the Internet protocols [6, 7, 8], and Unix network programming [10, 9] Serious students of Linux systems programming will want to study all of them. Tragically, Stevens died in 1999. His contributions will be greatly missed. The authoritative list of MIME types is maintained at ....

W. Richard Stevens. TCP/IP Illustrated: The Protocols, volume 1. Addison Wesley, 1994.

....of [42] Clock resolution. Many old computer systems (SVR4 386, BSD 386 V0.9, Linux x86 kernel V2.0) do not provide a time resolution higher than 10ms, either because they do not have better clock hardware or because their system times are generated by a 10 ms periodic interrupt, as explained in [43] Appendix B. This can seriously distort the measurement values. One solution is to measure the time every N loops and divide the time by N . N must be large enough so that the measured time is larger than the clock resolution. Most new micro processors such as Pentiums, Ultra )Sparcs and SGI MIPS ....

....One solution is to measure the time every N loops and divide the time by N . N must be large enough so that the measured time is larger than the clock resolution. Most new micro processors such as Pentiums, Ultra )Sparcs and SGI MIPS have special hardware (e.g. the Intel 8253 clock register [43]) that provide 1 s resolution, which is sucient for any practical Internet performance measurement. Therefore, we are not concerned with clock resolution problems in this paper. A related problem is timer resolution, which controls the granularity of how often an interrupt occurs. 18] nds that ....

W. R. Stevens. TCP/IP Illustrated: The Protocols, volume 1. Addison-Wesley, Reading, Massachusetts, December 1993.

....utldump, and lossy router) is created to later verify and analyze the results. 4 Lossy Router is an IP gateway that randomly drops certain packets according to a specified loss model and loss rate. 5 utldump is a modified version of the tcpdump program written by Jacobson, Leres, and McCanne [17]. tcpdump puts the network interface card into promiscuous mode so it captures every packet going across the wire. Then by applying a user specified filter (e.g. host names, port number, etc. packets of interest to the user are displayed. utldump performs the same job for UTL (Universal ....

W. Stevens. TCP/IP Illustrated: The Protocols, volume 1. Addison-Wesley, Reading, MA, 1994. 18

....layer protocol, ffl creates a utldump process on merlot to observe the traffic on the network, ffl sets the lossy router to simulate a specific loss rate, and 4 See Section 5.2.2. 5 utldump is a modified version of the tcpdump program written by Van Jacobson, Craig Leres, and Steven McCanne [148]. tcpdump puts the network interface card into promiscuous mode so that every packet going across the wire is captured. Then by applying a user specified filter (e.g. host names, port number, etc. packets that are interest to the user are displayed. utldump performs the same job for UTL ....

....and GIF89a s performance over S2E. In Experiment 10, we use protocols X2 and SP2 for GIFNCa and GIF89a, respectively. 9 ping was written by Mike Muuss to test whether another host is reachable. The program sends an ICMP echo request message to a host, expecting an ICMP echo reply to be returned [148]. 172 6.3.5.1 Experiment 9 In this experiment, we investigate the progressive display advantage of GIFNCa over GIF89a when images are transmitted over the actual Internet by using a transport layer protocol with flow control and no slow start. This experiment involves downloading the aircraft ....

W. Stevens. TCP/IP Illustrated: The Protocols, volume 1. Addison-Wesley, Reading, MA, 1994.

....is supported by the Australian Research Council under grants S499705, A49930405 and S4005343. 1 Introduction We have proposed in [10] a duality model of flow control 1 from which a basic algorithm is derived to maximize aggregate source utility. The significance of the model is that TCP Reno [5, 12] and Vegas [3] with or without random marking [4, 2] can all be interpreted within this model as approximately carrying out the basic algorithm to maximize aggregate utility, different schemes merely corresponding to different choices of utility functions; see [9] The model thus provides a ....

W. Stevens. TCP/IP illustrated: the protocols, volume 1. Addison--Wesley, 1999. 15th printing. 4

.... control is posed as an optimization problem in [6] where the objective is to maximize aggregate source utility (problem (5 6) in Section 3 below) It is solved using a penalty function approach in [8, 9] and a duality approach in [11] see also [12] The works [7, 9, 10] suggest that TCP Reno [15] can be interpreted in this optimization framework as a distributed algorithm to solve the maximization problem with a specific utility function. REM is originally proposed in [1] as a practical implementation of the dual algorithm of [11] This dual algorithm consists of a link subalgorithm and a ....

W. Stevens. TCP/IP illustrated: the protocols, volume 1. Addison--Wesley, 1999. 15th printing. 19

....not normally found in other transmission media, which adversely impact TCP performance. These are (i) long propagation delay, ii) a large bandwidth delay product, and (iii) bandwidth asymmetry. The congestion control (Slow Start) and congestion avoidance (multiplicative decrease) procedures [10 11] were first introduced by the Tahoe TCP implementation. This allows a TCP transmitter to transmit a limited number of bytes determined by the smallest value of the receiver advertised window or the congestion window (cwnd) The algorithm also uses a variable to keep the threshold value of the ....

....congestion avoidance (linear increase) phase. The high delay of a satellite link significantly slows the growth of the congestion window. Each ACK is delayed by the round trip time over the satellite link, and only when received, does it allow the sender to increase the cwnd by one segment (MSS [11]) during the slow start phase, and by a fraction of the MSS (equivalent to one MSS per round trip delay) during the linear increase phase. The Slow Start procedure [10] was introduced to achieve network stability under Internet congestion. It is however, known to be unduly conservative [11] ....

[Article contains additional citation context not shown here]

W. R. Stevens, TCP/IP Illustrated: The Protocols, Vol:1, Addison Wesley, New York, 1994.

....dgrams arriving on Loc IP p all dgrams sent to Loc port Figure 1. Specifying the four tuple implementations, including ours, implicitly set the remote address to loopback when connect( is invoked with a wildcard remote address. Figure 1 shows possibilities for values in the four tuple [10]. A p indicates a specific value while, indicates wildcard. The order of the three rows is the order used in demultiplexing a received datagram to a corresponding socket. The first row with most specific binding is tried first, and the last row with least specific binding is tried last. ....

R. Stevens. TCP/IP Illustrated: The Protocols, volume 1. Addison-Wesley, Reading, MA, USA, 1994.

....simple topology, see of Fig. 1, consisting of a bottlenecked link connecting m RTP based senders and receivers, n TCP based FTP connections and k TCP connections carrying WWW traffic. The TCP connections are based on the Reno TCP specifications with fast retransmission and fast recovery algorithms [13]. The LDA and FTP sources are modeled as greedy sources that always have data to send at the highest possible rate as allowed by the congestion control mechanism during the entire simulation time. The WWW servers are modeled as on off processes with the on period lasting for the time needed to ....

W. Richard Stevens, TCP/IP illustrated: the protocols, vol. 1, AddisonWesley, Reading, Massachusetts, 1994.

....When there is no buffer space available, the base station advertises a window size of zero, and the source refrains from transmitting any more data until the window size is increased. The base station will increase the window size only if it can accept at least one full TCP segment [11]. For WTCP and Snoop, the base station buffer cannot become full as the transport connection is end to end. For an endto end connection, the number of unacknowledged segments sent out by the source is bounded by the transmission window. Since the maximum window size is chosen to be smaller than ....

W. Richard Stevens. TCP/IP Illustrated The Protocols. Addison-Wesley, Reading, MA, USA, 1994.

....yields lower throughput. We believe that there is an optimal frame length that yields the best throughput under given conditions in wireless links. This procedure requires computation of the optimal frame length for the LBER and the HBER state in a way similar to the slow start TCP IP al gorithm [Ste94] This implies the use of different frame lengths for the LBER and the HBER state. 3.4 Pre emptive Retransmissions For loss sensitive traffic, it is possible that a frame may need to be retransmitted more than once due to a high error condition in the wireless channel. The pre emptive ....

R. Stevens. TCP/IP Illustrated: The Protocols, volume 1. Addison-Wesley Publishing Company, 1994.

....Scaling The biggest problem seen in TCP over ATM implementations is a limitation of the TCP protocol itself. This problem is not specific to ATM, but manifests itself whenever TCP is used over a network with a large bandwidth delay product. Such a network is known as a Long Fat Network, or LFN [18]. TCP is a sliding window protocol, so the sending and receiving hosts must first negotiate a window size, specified as a number of bytes, before transmitting any data. Subsequently, the sender cannot send more than a single window of data before receiving an acknowledgment from the receiver. ....

....bytes 1 : i Gamma 1 have been received. The TCP sender will then retransmit bytes i : n, even though the only bytes missing are i : j. The bandwidth used to retransmit bytes j 1 : n is entirely wasted. On LFNs, this effect causes a particularly catastrophic loss of efficiency [18]. Selective acknowledgment (SACK) allows the TCP receiver to report non sequential data that it has received, preventing wasted bandwidth caused by unnecessarily retransmitted data. RFC 1072 [22] first proposed a SACK extension to TCP, but the feature was omitted from RFC 1323 [21] because the ....

[Article contains additional citation context not shown here]

W. R. Stevens. TCP/IP Illustrated---The Protocols. Addison--Wesley, Reading, MA, 1994.

....we increase x by 0:1. This allows the system to nd a fraction that successfully controls temperature without excessively reducing performance. Though simple, this algorithm is ecient and conceptually similar to algorithms found in other elds, e. g, the congestion avoidance algorithm in TCP IP [16]. By dynamically updating the activity of hot processes, we can precisely adjust the highest CPU level the system can safely support. The system also automatically reacts to newly created processes or dead processes by regularly updating the list of hot processes and sharing the maximum allowed ....

W. Stevens. TCP/IP Illustrated - The Protocols, chapter 21, pages 299-322. Addison-Wesley, Reading, MA, USA, 1994.

....on the usage of RED routers. Based on results achieved in [6] the minimum drop threshold was set to 0.4 and the maximum to 0.9 of the route buffer. The maximum buffering delay was set to 0.3 seconds. The TCP sender is modulated based on the Reno TCP model with fast recovery and retransmission [16]. Both the TCP and RTP senders transmit data packets of 1000 Bytes and always have data to send. 4.2.1 Statical Multiplicative Decrease Factor The work done in [4] shows analytically, that using a constant multiplicative decrease factor results in a fair bandwidth distribution among competing ....

W. R. Stevens. TCP/IP illustrated: the protocols, volume 1. Addison-Wesley, Reading, Massachusetts, 1994.

....of LDA we used a simple topology, see Fig. 1, consisting of a bottlenecked link connecting m RTP based senders and receivers, n TCP based FTP connections and k WWW servers. The TCP connections were based on the Reno TCP specifications with fast retransmission and fast recovery algorithms [15]. The LDA and FTP sources were modeled as greedy sources that always had data to send at the rate allowed by the congestion control mechanism. The WWW servers generated short TCP connections with each connection lasting for the time required to carry a number of packets drawn from a Pareto ....

W. Richard Stevens, TCP/IP illustrated: the protocols, vol. 1, AddisonWesley, Reading, Massachusetts, 1994.

.... requiring any changes in the network [5] However, such an approach requires the exchange of state information between the source end system and the network as it is the case for the available bit rate service of ATM [17] or between the sender and the receivers as its is the case for TCP [23] and the scheme presented here. As these state information have to traverse at least a part of the network, the source will not be able to change its sending rate immediately after a change in the Sisalem: Fairness of Adaptive Multimedia Applications 2 congestion state of the network. In Sec. 2, ....

....and the differences in the way TCP and the LBA algorithm react to congestion in the network when TCP connections share a bottlenecked link with connections using an adaptive scheme. As a simulation topology we used the topology depicted in Fig. 6. The TCP source was based on 4. 3BSD Reno TCP [23]. Router Router 1000 Km TCP Sender Adaptive Sender 10 Mb s Adaptive Sender Figure 6: Test configuration for the interaction of the LBA algorithm and TCP The results obtained from our simulations, see Fig. 7, suggest that the TCP connection, actually, gets starved and receives on the ....

W. R. Stevens. TCP/IP illustrated: the protocols, volume 1. Addison-Wesley, Reading, Massachusetts, 1994.

....ALS connections sharing a bottleneck router with some TCP connections. Fig. 2 shows a simple network topology with n TCP connections sharing a link of b kb s with m ALS connections. The TCP connections are based on the Reno TCP specifications with fast retransmission and fast recovery algorithms [28]. Both the ALS and TCP sources are persistent sources. That is, they always have data to send at the highest allowed transmission rate. The router used in the simulations as well in the simulations in the next sections uses random early detection (RED) 11] for buffer management. A RED router ....

W. R. Stevens. TCP/IP illustrated: the protocols, volume 1. Addison-Wesley, Reading, Massachusetts, 1994.

....still face different problems in terms of complexity, scalability and applicability. The second approach is to adapt the amount of data entering the network in accordance with the network congestion situation. This has been widely used in the Internet through the transport control protocol (TCP) [22] used for data applications such as WWW and FTP transports. With TCP, the sender increases its transmission rate by an additive amount while no losses are observed. After observing a loss the rate is reduced by a multiplicative factor. To avoid overloading the network, non TCP flows, e.g. ....

W. R. Stevens. TCP/IP Illustrated: The Protocols, volume 1. Addison-Wesley, Reading, Massachusetts, USA, Dec. 1993. 20

.... the timestamp of an incoming sender report (T sender ) and the time the report 6 arrived at the receiver as indicated by the receiver s local clock (T receiver ) 2 = T receiver Gamma T sender (8) A smoothed round trip delay (t RTT ) can then be estimated similar to the approach used with TCP [28] t RTT = t RTT 0:125 Theta ( Gamma t RTT ) 9) However, for this one way measurement approach to deliver accurate delay estimations some kind of synchronization of the clocks among the receivers and sender is required. Additionally, Paxon [26] shows through extensive measurements in the ....

....capacities. Each router is shared between an MLDA stream and m TCP connections that have the same end toend propagation delay as that of the MLDA sender receiver pair. The TCP connections are modeled as FTP flows that always have data to send and last for the entire simulation time. A TCP Reno [28] model was used for simulating the congestion control behavior of TCP. The sender transmits packets of 1 kbytes and each router is a random early drop (RED) 31] router. A RED gateway detects incipient congestion by computing the average queue size. When the average queue size exceeds a preset ....

[Article contains additional citation context not shown here]

W. Richard Stevens, TCP/IP illustrated: the protocols, vol. 1, Addison-Wesley, Reading, Massachusetts, 1994.

....of data, the library transparently communicates multiple times until it has sent all of the data. Although a larger MSS (up to the size of the maximum buffer size of the underlying network) is often better, especially for local communication, the default non local size is normally only 536 bytes [52]. We therefore performed a throughput experiment with the MSS set to this size; this is advantageous to handlers, for there is more work to be done which is application independent and can thus be handled by the library ASH transparently to the application. For this experiment, we also decreased ....

R.W. Stevens. TCP/IP illustrated: the protocols, volume 1, chapter 18, page 237. Addison-Wesley Pub. Co., 1994.

....When there is no buffer space available, the base station advertises a window size of zero, and the source refrains from transmitting any more data until the window size is increased. The base station will increase the window size only if it can accept at least one full TCP segment [9]. Fixed Host (Source) Base Station Mobile Host (Sink) 100 ms 800 Kbps 10 ms 10 Mbps Figure 1. Simulated network. 3 Results and Discussion We measured the throughput of I TCP, Snoop, WTCP and TCP tahoe 6 , and the results are shown in Figure 2. Throughput is measured as the average number of ....

W.R. Stevens. TCP/IP Illustrated The Protocols. Addison-Wesley, Reading, MA, USA, 1994.

....not be very useful to draw conclusions about the cause of a packet loss. 4 Experiments We use a set of data collected by Vern Paxson [5] to study end to end Internet dynamics. Paxson collected four sets of data R 1 , R 2 , N 1 , and N 2 . R 1 and R 2 are traces collected with the tool traceroute [7] and were used to study routing. N 1 and N 2 are tcpdump[7] traces collected over 37 sites to study end to end packet dynamics. The measurements are extensively described in [5] We used the set N 2 which contains about 20,000 tcpdump traces for bulk transfers of 100 KBytes between 2 sites among ....

....a packet loss. 4 Experiments We use a set of data collected by Vern Paxson [5] to study end to end Internet dynamics. Paxson collected four sets of data R 1 , R 2 , N 1 , and N 2 . R 1 and R 2 are traces collected with the tool traceroute [7] and were used to study routing. N 1 and N 2 are tcpdump[7] traces collected over 37 sites to study end to end packet dynamics. The measurements are extensively described in [5] We used the set N 2 which contains about 20,000 tcpdump traces for bulk transfers of 100 KBytes between 2 sites among the 37 sites studied. For each transfer, N 2 contains the ....

W. R. Stevens. TCP/IP Illustrated: the protocols (v.1). Addison-Wesley, 1994.

No context found.

W. R. Stevens, TCP/IP Illustrated: The Protocols. Addison-Wesley, vol. 1.

No context found.

R. W. Stevens, TCP/IP Illustrated: the protocols. Addison-Wesley, 1994.

No context found.

W. R. Stevens, TCP/IP illustrated: the protocols, Vol. 1. Reading, Massachusetts: Addison-Wesley, 1994.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC