W. R. Stevens, TCP/IP Illustrated, Volume 1, Addison-Wesley, Reading, MA, November 1994.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....TCP compatibility, nonlinear algorithms, transport protocols, TCP, streaming media, Internet. I. INTRODUCTION The stability of the Internet to date has in large part been due to the congestion control and avoidance algorithms [1] implemented in its dominant transport protocol, TCP [2] [3]. Based on the principle of additive increase multiplicativedecrease (AIMD) 4] a TCP connection probes for extra bandwidth by increasing its congestion window linearly with time, and on detecting congestion, reducing its window multiplicatively by a factor of two. Under certain assumptions of ....

W. R. Stevens, TCP/IP Illustrated, Volume 1, Addison-Wesley, Reading, MA, November 1994.

....in the fixed net for supporting mobility. To analyze this overhead, we consider the traffic in the fixed net that is generated by or destined for a user, and compare the case where she is on a mobile node to the case where she is on a fixed node. Here, the fixed net may be an autonomous subsystem [St], or the Internet itself. The traffic pattern depends on whether the correspondents are fixed or mobile, and whether a mobile node is in its home cell; this is illustrated in Figure 1 for two cases. Altogether, there are 29 cases (using one particular way of counting) Our comparison is based ....

....collisions are frequent. Therefore p busy = M per host T packet 1 Gamma p noise : 1) A mobile node is likely to encounter wireless cells with varying bandwidths [FZ] and must adjust its rate of transmission attempts accordingly. One popular mechanism for doing so is the exponential backoff [MB, St], for which there is a constant T backoff so that, if a node finds the channel busy, it remains idle for a random period with mean 2 i Gamma1 T backoff after the ith attempt (i = 1; 2; The waiting time for the channel to be idle is therefore Twait = T backoff p busy (1 Gamma p busy ) ....

W.R. Stevens, TCP/IP Illustrated, Volume 1, Addison-Wesley (1994).

....provision of packet data services for applications like e mail, web browsing and mobile computing over wireless channels is gaining importance. Transport Control Protocol (TCP) is a reliable, end to end, transport protocol that is widely used to support applications like telnet, ftp, and http [1]. TCP was designed for wireline networks where the channel error rates are very low and congestion is the primary cause of packet loss [2] Since its original deployment, several modifications to TCP, including Reno, NewReno, and Vegas have been proposed and their performance analyzed in wireline ....

....acknowledged. The change in W (t) however, depends on the particular version of TCP and the congestion control process. Each time a new packet is transmitted, the transmitter starts a timer. If such timer reaches the round trip timeout value (derived from a round trip time estimation procedure [1]) before the packet is acknowledged, timeout expiration occurs and retransmission is initiated from the next packet after the last acknowledged packet. It is noted that the timeout values are set only in multiples of a timer granularity [1] The basic window adaptation procedure, common to all ....

[Article contains additional citation context not shown here]

W. R. Stevens, TCP/IP Illustrated, Volume 1, Addison Wesley, 1994.

....Since new data is generated in response to acknowledgements, roughly, the effective data transmission rate is given by the window size (in bytes) divided by the round trip delay in the connection. Thus window adaptation indirectly results in rate adaptation. For details on the TCP protocol see [Stevens 94] 8 The design approach behind TCP is in keeping with the general Internet philosophy of a simple network connecting intelligent terminals. TCP expects little from the network, and is itself a very complex protocol that runs in the end systems. Owing to its simplicity and its lack of ....

W.R. Stevens, TCP/IP Illustrated, Volume 1, Addison-Wesley, Reading, Mass., Nov. 1994. 18

....advances in the area of wireless communications and Internet, provision of data services for applications like e mail, web browsing, telnet, etc. over wireless is gaining importance. Transport Control Protocol (TCP) is a reliable end to end transport protocol in the Internet Protocol (IP) suite [1]. TCP is tuned to perform well in wireline networks where the channel error rates are very low and congestion is the primary cause of packet loss [2] However, when used over wireless channels which are characterized by high frame error rates (FER) the performance of TCP is severely affected ....

....TCP layer is no smaller than 536 bytes (default value) the trans mitted TCP segment size is no larger than 2047 bytes, and the advertised window size is no smaller than 2 MSS and no larger than 4 MSS. Successful transmission of each TCP segment is ensured through an acknowledgement based scheme [1]. The RLP data unit, called a frame, on the other hand, is about 20 bytes in length (equivalent to a fixed frame duration of 20 msec at a transmission rate of 9600 bps) Hence, a typical TCP segment transmission requires about 30 RLP frame transmissions. When transferring data, RLP is a pure ....

W. R. Stevens, TCP/IP Illustrated, Volume 1, Addison Wesley, 1994.

....advances in the area of wireless communications and internet, provision of data services for applications like e mail, web browsing, telnet, etc. over wireless is gaining importance. Transport Control Protocol (TCP) is a reliable end to end transport protocol in the Internet Protocol (IP) suite [1]. TCP is tuned to perform well in wireline networks where the channel error rates are very low and congestion is the primary cause of packet loss [2] However, when used over wireless channels which are characterized by high frame error rates (FER) the performance of TCP is severely affected ....

....TCP layer is no smaller than 536 bytes (default value) the transmitted TCP segment size is no larger than 2047 bytes, and the advertized window size is no smaller than 2 MSS and no larger than 4 MSS. Successful transmission of each TCP segment is ensured through an acknowledgement based scheme [1]. The RLP data unit, called a frame, on the other hand, is about 20 bytes in length (equivalent to a fixed frame duration of 20 msec at a transmission rate of 9600 bps) Hence, a typical TCP segment transmission requires about 30 RLP frame transmissions. When transferring data, RLP is a pure ....

W. R. Stevens, TCP/IP Illustrated, Volume 1, Addison Wesley, 1994.

....(TCP) Pos81b] and User datagram Protocol (UDP) Pos80] Both are described in more detail in the following sections (3.1 and 3. 2) IP provides for transmitting blocks of data, called datagrams or packets, from sources to destinations [Pos81a] It is a connectionless, unreliable protocol [Ste98] Connectionless meaning that each packet is treated independently, and packets can arrive in a di erent order to what they were sent. Unreliable referring to the fact that there is no guarantee that each packet will eventually get to it s nal destination. The IP header contains information ....

....layer protocol can provide services such as error correction and detection, congestion avoidance and connection establishment. The two main protocols, as mentioned earlier, are TCP and UDP. 3. 1 Transmission Control Protocol TCP provides a connection oriented, reliable, byte stream service [Ste98] Before any data can be sent between two machines a connection must be established. One machine will send a request for a connection specifying the port number and machine it wishes to connect to. The responding machine will acknowledge this request and a connection will be established. This ....

W. Stevens. TCP/IP Illustrated, volume 1. Addison-Wesley Publishing Company, September 1998.

....advances in the area of wireless communications and internet, provision of data services for applications like e mail, web browsing, telnet, etc. over wireless is gaining importance. transport control protocol (TCP) is a reliable end to end transport protocol in the internet protocol (IP) suite [1]. TCP is tuned to perform well in wireline networks where the channel error rates are very low and congestion is the primary cause of packet loss [2] However, when used over wireless channels which are characterized by high frame error rates (FER s) the performance of TCP is severely affected ....

....CHOCKALINGAM AND BAO: PERFORMANCE OF TCP RLP PROTOCOL STACK 29 transmitted TCP segment size is no larger than 2047 bytes, and the advertised window size is no smaller than 2 MSS and no larger than 4 MSS. Successful transmission of each TCP segment is ensured through an acknowledgment based scheme [1]. The RLP data unit, called a frame, on the other hand, is about 20 bytes in length (equivalent to a fixed frame duration of 20 ms at a transmission rate of 9600 bps) Hence, a typical TCP segment transmission requires about 30 RLP frame transmissions. When transferring data, RLP is a pure ....

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

....related parallel programming models have become widely accepted. However, the execution of parallel applications on distributed systems has been hampered by the high communication overhead. For parallel computing using Network of Workstations (NOWs) it is obvious that the TCP IP protocol layers [12] of the communication network are the bottleneck in communication and introduce a major delay to the message passing in a parallel application [1] In the current implementation of the TCP IP protocol layers, messages are handled in a First Come FirstServer (FCFS) manner. Thus, under a heavy ....

G.R. Wright and W.R. Stevens, TCP/IP Illustrated, Volume2, The Implementation, AddisonWesley, Reading, MA, 1995.

....data call. The PPP layer is used for initial call establishment and to negotiate initial optional link capabilities like maximum PPP frame size [2] Both IP and PPP layers add fixed number of overhead bytes (e.g. 16 byte uncompressed IP header or 3 byte VJ compressed IP header, 4 byte PPP header [8]) Since the bulk throughput and energy efficiency performance of the stack during data transfer phase mainly depends on the ARQ mechanism at the LL, we focus mainly on UDP and LL. In particular, we ignore IP and PPP layers in our model as, from a performance view point, they will merely add their ....

W. R. Stevens, TCP/IP Illustrated, Volume 1, Addison Wesley, 1994.

....not necessarily imply fairness relative to TCP performance, especially for drop tail bottleneck gateways. 1 Introduction The stability of the Internet to date has in large part been due to the congestion control and avoidance algorithms [15] implemented in its dominant transport protocol, TCP [28, 34]. Based on the principle of additive increase multiplicativedecrease (AIMD) 6] a TCP connection probes for extra bandwidth by increasing its congestion window linearly with time, and on detecting congestion, reducing its window multiplicatively by a factor of two. Under certain assumptions of ....

STEVENS, W. R. TCP/IP Illustrated, Volume 1. AddisonWesley, Reading, MA, Nov 1994.

....sender, so that the network is continuing to pass timing signals to the sender indicating the rate at which packets are arriving at the receiver. Only when the repair has been completed does the sender drop its window to the ssthresh value as part of the transition to congestion avoidance mode [8] . The other signal of packet loss is a complete cessation of any ACK packets arriving to the sender. The sender cannot wait indefinitely for a delayed ACK, but must make the assumption at some point in time that the next unacknowledged data segment must be retransmitted. This is managed by the ....

Stevens, W. R., TCP/IP Illustrated, Volume 1, Addison-Wesley, 1994.

....situation is to reduce the load in an attempt to match the available capacity of this bottleneck link. Protocols that follow this approach and use some form of congestion control are referred to as responsive. The most common example of a responsive protocol is the Transport Control Protocol (TCP) [1]. Those protocols that fail to detect or choose to ignore congestion by simply maintaining (or even increasing ) their load are referred 1 Here we use the term flow simply as a convenient way to designate the packets traveling between a source and destination. In the IP context, a flow is the ....

....In the IP context, a flow is the collection of packets having a common addressing 5 tuple of source, source port, destination, destination port, and transport protocol type. 2 to as unresponsive. The most common example of an unresponsive protocol is the Unreliable Datagram Protocol (UDP) [1]. The most widely advocated responsive mechanism is to decrease the generated load geometrically when congestion is detected and increase load linearly in order to probe for available capacity when the congestion subsides [2] Because this approach was first shown to be effective in TCP, those ....

W. R. Stevens, TCP/IP Illustrated, Vol. 1: The Protocols. Addison-Wesley, Reading, Mass. 1994.

....etc. are gaining increased attention due to rapid advances in the areas of wireless communications and the Internet. Transmission Control Protocol (TCP) is an end to end transport protocol in the Internet Protocol (IP) suite, which is widely used in popular applications like telnet, ftp and http [1]. TCP guarantees reliable, in sequence delivery of packets. TCP performance gets severely affected when used on channels which are typically characterized by high error rates (e.g. wireless channels) 2] Tahoe and Reno versions of TCP are commonly used implementations. 1] Several other ....

....telnet, ftp and http [1] TCP guarantees reliable, in sequence delivery of packets. TCP performance gets severely affected when used on channels which are typically characterized by high error rates (e.g. wireless channels) 2] Tahoe and Reno versions of TCP are commonly used implementations. [1]. Several other enhancements to TCP, including New Reno, SACK, and Vegas, have been proposed to improve TCP performance [3] 4] Such enhanced versions have been shown to provide better performance than Tahoe in wireline networks [4] However, in [5] it has been shown that TCP enhancements like ....

W. R. Stevens, TCP/IP Illustrated, Volume 1, Addison Wesley, 1994.

....night . Finally, we present our conclusions in section 7. 2. Test Methodology In this section, we will first discuss our two test tools, ping and ttcp. We also provide a short discussion of how our tests were setup and performed. 2. 1 Ping and TTCP Our tests are based on the FreeBSD ping [PING] [STEVENS94] program and the public domain ttcp tool [TTCP] We wanted a ping that was flood ping capable, and had high temporal granularity in terms of its response. This in theory, would allow us to make reasonable measurements both in terms of high speed gigabit ethernet connections, and over relatively ....

....TCP didn t really show that much slowdown with 32k windows between the (e.3) OHSU to Pittock, and (e.5) PSU to OHSU links. This is perhaps not news, but it is comforting to notice that different delays (which are longer delays) did not seem to change the result appreciably. 5 For example, see [STEVENS94], pp. 344 for his discussion of the bandwidth delay product. 6. Things That Go Bump In The Night In this section we are briefly going to review some miscellaneous tests, some of which were host oriented and involved non BSD operating systems. One of the more interesting aspects of the entire ....

Stevens, Richard W., TCP/IP Illustrated, Volume 1, chapter 7, 1994, Prentice-Hall, pp.8590

....well, one should note that a topic of on going research is the enhancement of closed loop controls to improve the throughput that is indeed realized in various contexts. The implemented changes in TCP of congestion avoidance, fast retransmit, and fast recovery have this aim, see for example [15]. Further changes are also being proposed, such as explicit congestion notification, 16, 17] The design of the feedback control for the Available Bit Rate transfer capability in ATM received much attention at the ATM Forum, an industry forum; for a summary see Fendick [18] Also, of recent ....

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

....programming languages and related parallel programming models have become widely accepted. However, the execution of parallel applications on distributed systems has been hampered by the high communication overhead. For parallel computing using NOW, it is obvious that the TCP IP protocol layers [1] of the communication network are the bottleneck in communication and introduce a major delay to the message passing in a parallel application [2] In the current implementation of the TCP IP protocol layers, messages are handled in an FIFO manner. Thus, under a heavy network traffic load ....

.... Gamma 1 ) 1= Gamma 1 ) level 2, t d2 1= 1 2 ) Gamma 1 ) Gamma 1 Gamma 2 ) Gamma 1 ) Gamma 1 Gamma 2 ) Table 1: Average Total Delay per Packet Using Priority Policies to the interface layer. Thus, a package independently enters the queue at the interface layer [1]. To model the message delay, a message queue is assumed as follows: The queue policy is either FIFO or priority M M 1. The total delay of a message at a node consists of the queuing time and the transmission time of the message. A message consists of packages, and the packages in the ....

G.R. Wright and W.R. Stevens, TCP/IP Illustrated, Volume2, The Implementation, Addison-Wesley, Reading, MA, 1995.

....causes many triple duplicate ACKS and TCP timeouts which severely impair TCP throughput. All handoffs cause the loss of one or more packets. With the optimized handoff protocol on the other hand, no packets are lost due to handoffs at all. The periodic 1 For a more thorough discussion see e.g. Ste94] 4.1 Evaluation of the Optimized Handoff Protocol 19 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 Throughput (KBit s) Time (s) a) Standard Handoff Mechanism 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 Throughput (KBit s) Time (s) b) Optimized Handoff ....

Richard W. Stevens. TCP/IP Illustrated, Volume 1. The Protocols. Addison-Wesley Publishing Company, 1994.

....N. In this paper, we discuss how to use this universal addressing in a number of protocols for routing data messages to mobile computers, for establishing and maintaining mobile vir tual circuits, and for supporting mobile groups. 1. Introduction We consider a system, like the Internet [2] and [6], that consists of several computer networks. The computers in each network are classified into hosts and routers. Each network has one or more hosts and one or more routers. For convenience, we assume that each host is attached to exactly one network, and each router is attached to two or more ....

W. R. Stevens, TCP/IP Illustrated, Volume 1, Addison-Wesley, 1994.

.... to recover from single packet losses by waiting for three duplicate ACKs, retransmitting the lost packet (fast retransmit) and then resuming normal sends with half the window size (fastrecovery) Details of TCP s mechanisms can be found in the paper by Jacobson et al. 16] and a book by Stevens [39]. 2.3.2 TCP friendliness In the previous section we saw the principles of congestion control that are employed by TCP streams. While the careful design affords TCP the stability and flexibility to effectively use the Internet, it leaves TCP vulnerable to bandwidth stealing by UDP based flows or ....

....of 1.5 Mbps, which is representative of a T1 link. We use a 1,500 byte packet size, which is a common size of packets seen in the Internet [7] A maximum receiver window of 10 packets (15,000 bytes) is used which is near the higher end of the default values used for typical TCP implementations [39]. We assume that all the data transfers are unidirectional and therefore set the ACK size to 40 bytes, which is the size of a TCP ACK with no payload. The source and destination hosts connect to the bottleneck link with a 10 Mbps link which represents a local area network. Previous simulation ....

[Article contains additional citation context not shown here]

W.R. Stevens. TCP/IP Illustrated, Volume 1. Addison Wesley, 1994.

....the performance of a client server strategy with an MA strategy [8] Weather monitoring and satellite data processing applications have been developed with the client server model and MA model. Performance is compared using collected measurements. 1 For details about TCP IP trac, see Reference [19]. 4 Sahai and Morin performed similar work in comparing a client server approach with an MA approach [18] They measured bandwidth utilization and response time of network management tasks consisting of getting a varying number of samples (using SNMP get requests) on a varying number of ....

W.R. Stevens. TCP/IP Illustrated, Volume 1. Addison-Wesley Publishing Company, Inc., Reading, Massachusetts, professional computing series edition, 1994.

....quickly answer what if questions, or evaluate the effect of modifying a particular protocol, or network parameter. In this paper, we develop models and analyses for studying the bulk throughput of four versions of TCP; namely, OldTahoe (the original protocol from [7] Tahoe, Reno, and NewReno [15], 6] We assume a local network scenario, in which a host on a local area network (LAN) is transmitting bulk data to a mobile host connected to the LAN by a wireless link. Our models incorporate important aspects, such as, slow start and congestion avoidance, coarse timers, fastretransmit, and ....

....skip Section V; the notation and terminology used in Section VI are defined before Section V. Finally, some conclusions are summarised in Section VII. II. The TCP Protocol We model only the data transfer part of TCP. Details of the TCP protocol can be found in the various Internet RFCs; see also [15]. The versions of the TCP protocol that we model and analyse all assume the same receiver process. The TCP receiver: accepts packets out of sequence number order, buffers them in a TCP buffer, and delivers them to its TCP user in sequence. Since the receiver has a finite resequencing buffer, it ....

[Article contains additional citation context not shown here]

W.R. Stevens, TCP/IP Illustrated, Volume 1, Addison-Wesley, Reading, Mass., Nov. 1994.

....byte headers. IP usually requires a link layer to handle hop by hop connections, although it is possible to use IP without this layer. Two principle transport layer protocols use IP. UDP [6] like its underlying network layer, is a best e ort, connectionless, datagram delivery service, while TCP [7] develops a reliable, connection oriented, data stream protocol that uses IP datagrams. 1.2 Asynchronous Transfer Mode (ATM) ATM [5] is a cell based, link layer switching and multiplexing technology designed to be used as a general purpose, connection oriented transfer mode for a wide range of ....

....protocol that provides 2 byte source port and destination port elds. These, if amalgamated as a single 4 byte eld, could be used to carry the ATM cell header (without its 1 byte HEC eld) However, this does nothing to solve the requirement that sequencing information must be included. TCP [7] is a protocol able to carry the required information. However, its slow start and complex re transmission algorithms make it much too cumbersome for this purpose. The alternative is to de ne a new IP option and to use this to hold the required header information. I decided to explore this option ....

W. Richard Stevens. TCP/IP Illustrated, Volume 1, chapter 17. Addison{ Wesley, 1994.

....e ort, connectionless, datagram delivery service at the network layer, using variable length datagrams with 20 byte headers. IP usually requires a link layer to handle hop by hop connections, although it is possible to use IP without this layer. Two principle transport layer protocols use IP. UDP [6], like its underlying network layer, is a best e ort, connectionless, datagram delivery service, while TCP [7] develops a reliable, connection oriented, data stream protocol that uses IP datagrams. 1.2 Asynchronous Transfer Mode (ATM) ATM [5] is a cell based, link layer switching and multiplexing ....

....of the IP network. The requirement that ATM cells arrive in the same order in which they were transmitted means that some form of sequencing must be added. The information in ATM signalling cells and ATM resource management cells must also be carried transparently across the IP network. UDP [6] is a simple, best e ort, transport level protocol that provides 2 byte source port and destination port elds. These, if amalgamated as a single 4 byte eld, could be used to carry the ATM cell header (without its 1 byte HEC eld) However, this does nothing to solve the requirement that ....

W. Richard Stevens. TCP/IP Illustrated, Volume 1, chapter 11. Addison{ Wesley, 1994.

No context found.

W. Stevens. TCP/IP Illustrated, Vol. 1, The Protocols. Addison-Wesley, 1997.

First 50 documents  Next 50

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