Classifications

• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L1/00—Arrangements for detecting or preventing errors in the information received
  • H04L1/12—Arrangements for detecting or preventing errors in the information received by using return channel
  • H04L1/16—Arrangements for detecting or preventing errors in the information received by using return channel in which the return channel carries supervisory signals, e.g. repetition request signals
  • H04L1/18—Automatic repetition systems, e.g. Van Duuren systems
  • H04L1/1829—Arrangements specially adapted for the receiver end
  • H04L1/1832—Details of sliding window management
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04W—WIRELESS COMMUNICATION NETWORKS
  • H04W28/00—Network traffic management; Network resource management
  • H04W28/02—Traffic management, e.g. flow control or congestion control
  • H04W28/0289—Congestion control
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L1/00—Arrangements for detecting or preventing errors in the information received
  • H04L1/12—Arrangements for detecting or preventing errors in the information received by using return channel
  • H04L1/16—Arrangements for detecting or preventing errors in the information received by using return channel in which the return channel carries supervisory signals, e.g. repetition request signals
  • H04L1/18—Automatic repetition systems, e.g. Van Duuren systems
  • H04L1/1867—Arrangements specially adapted for the transmitter end
  • H04L1/187—Details of sliding window management
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L47/00—Traffic control in data switching networks
  • H04L47/10—Flow control; Congestion control
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L47/00—Traffic control in data switching networks
  • H04L47/10—Flow control; Congestion control
  • H04L47/12—Avoiding congestion; Recovering from congestion
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L47/00—Traffic control in data switching networks
  • H04L47/10—Flow control; Congestion control
  • H04L47/26—Flow control; Congestion control using explicit feedback to the source, e.g. choke packets
  • H04L47/263—Rate modification at the source after receiving feedback
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L47/00—Traffic control in data switching networks
  • H04L47/10—Flow control; Congestion control
  • H04L47/30—Flow control; Congestion control in combination with information about buffer occupancy at either end or at transit nodes
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L47/00—Traffic control in data switching networks
  • H04L47/10—Flow control; Congestion control
  • H04L47/32—Flow control; Congestion control by discarding or delaying data units, e.g. packets or frames
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
  • H04L69/16—Implementation or adaptation of Internet protocol [IP], of transmission control protocol [TCP] or of user datagram protocol [UDP]
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
  • H04L69/16—Implementation or adaptation of Internet protocol [IP], of transmission control protocol [TCP] or of user datagram protocol [UDP]
  • H04L69/161—Implementation details of TCP/IP or UDP/IP stack architecture; Specification of modified or new header fields
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
  • H04L69/16—Implementation or adaptation of Internet protocol [IP], of transmission control protocol [TCP] or of user datagram protocol [UDP]
  • H04L69/163—In-band adaptation of TCP data exchange; In-band control procedures
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
  • H04L69/16—Implementation or adaptation of Internet protocol [IP], of transmission control protocol [TCP] or of user datagram protocol [UDP]
  • H04L69/168—Implementation or adaptation of Internet protocol [IP], of transmission control protocol [TCP] or of user datagram protocol [UDP] specially adapted for link layer protocols, e.g. asynchronous transfer mode [ATM], synchronous optical network [SONET] or point-to-point protocol [PPP]
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04W—WIRELESS COMMUNICATION NETWORKS
  • H04W24/00—Supervisory, monitoring or testing arrangements
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04W—WIRELESS COMMUNICATION NETWORKS
  • H04W28/00—Network traffic management; Network resource management
  • H04W28/02—Traffic management, e.g. flow control or congestion control
  • H04W28/0231—Traffic management, e.g. flow control or congestion control based on communication conditions
  • H04W28/0242—Determining whether packet losses are due to overload or to deterioration of radio communication conditions
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04W—WIRELESS COMMUNICATION NETWORKS
  • H04W28/00—Network traffic management; Network resource management
  • H04W28/02—Traffic management, e.g. flow control or congestion control
  • H04W28/0273—Traffic management, e.g. flow control or congestion control adapting protocols for flow control or congestion control to wireless environment, e.g. adapting transmission control protocol [TCP]
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04W—WIRELESS COMMUNICATION NETWORKS
  • H04W28/00—Network traffic management; Network resource management
  • H04W28/02—Traffic management, e.g. flow control or congestion control
  • H04W28/10—Flow control between communication endpoints
  • H04W28/14—Flow control between communication endpoints using intermediate storage
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04W—WIRELESS COMMUNICATION NETWORKS
  • H04W8/00—Network data management
  • H04W8/02—Processing of mobility data, e.g. registration information at HLR [Home Location Register] or VLR [Visitor Location Register]; Transfer of mobility data, e.g. between HLR, VLR or external networks
  • H04W8/04—Registration at HLR or HSS [Home Subscriber Server]
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04W—WIRELESS COMMUNICATION NETWORKS
  • H04W80/00—Wireless network protocols or protocol adaptations to wireless operation
  • H04W80/06—Transport layer protocols, e.g. TCP [Transport Control Protocol] over wireless
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L1/00—Arrangements for detecting or preventing errors in the information received
  • H04L1/12—Arrangements for detecting or preventing errors in the information received by using return channel
  • H04L1/16—Arrangements for detecting or preventing errors in the information received by using return channel in which the return channel carries supervisory signals, e.g. repetition request signals
  • H04L1/1607—Details of the supervisory signal

Definitions

• This invention pertains generally to routing data in wireless networks, and more particularly to improving TCP performance in wireless networks by distinguishing congestion versus random loss.
• router-based support for TCP congestion control can be provided through RED gateways, a solution in which packets are dropped in a fair manner (based upon probabilities) once the router buffer reaches a predetermined size.
• an Explicit Congestion Notification (ECN) bit can be set in the packet header, prompting the source to slow down.
• ECN Explicit Congestion Notification
• current TCP implementations do not support the ECN method.
• An approach has also been proposed that prevents TCP sources from growing their congestion window beyond the bandwidth delay product (BWDP) of the network by allowing the routers to modify the receiver's advertised window field of the TCP header in such a way that TCP does not overrun the intermediate buffers in the network.
• BWDP bandwidth delay product
• End-to-end congestion control approaches can be separated into three categories: using rate control, looking at changes in packet round-trip time (RTT), and modifying the source and/or receiver to return additional information beyond what is specified in the standard TCP header.
• RTT packet round-trip time
• a problem with rate-control and relying upon RTT estimates is that variations of congestion along the reverse path cannot be identified. Therefore, an increase in RTT due to reverse-path congestion or even link asymmetry will affect the performance of these algorithms in an adverse manner.
• the window size could be decreased (due to increased RTT) resulting in decreased throughput; in the case of rate-based algorithms, the window could be increased in order to bump up throughput, resulting in increased congestion along the forward path.
• the DUAL algorithm uses a congestion control scheme that examines the
• RTT variation as the indication of delay through the network.
• the algorithm keeps track of the minimum and maximum delay observed to estimate the maximum queue size in the bottleneck routers and keep the window size such that the queues do not fill and thereby cause packet loss.
• RFC 1185 uses the TCP Options to include a timestamp, in every data packet from sender to receiver in order to obtain a more accurate RTT estimate. The receiver echoes this timestamp in any acknowledgment (ACK) packet and the round-trip time is calculated with a single subtraction. This approach encounters problems when delayed ACKs are used because it is unclear to which packet the timestamp belongs.
• RFC1185 suggests that the receiver return the earliest timestamp so that the RTT estimate takes into account the delayed ACKs, as segment loss is assumed to be a sign of congestion, and the timestamp returned is from the SN which last advanced the window. When a hole is filled in the sequence space, the receiver returns the timestamp from the segment which filled hole.
• the downside of this approach is that it cannot provide accurate timestamps when segments are lost.
• Tri-S scheme Two notable rate-control approaches are the Tri-S scheme and. TCP Vegas.
• the Tri-S algorithm is a rate-based scheme that computes the achieved throughput by measuring the RTT for a given window size (which represents the amount of outstanding data in the network). It compares the throughput for a given window and then for the same window increased by one segment. If the throughput is less than one-half that achieved with the smaller window, they reduce the window by one segment.
• TCP Vegas tries to prevent congestion by estimating the expected throughput and then adjusting the transmission window to keep the actual observed throughput close the expected value.
• Another approach is bandwidth probing.
• two back-to-back packets are transmitted through the network and the interarrival time of their acknowledgment packets is measured to determine the bottleneck service rate (the conjecture is that the ACK spacing preserves the data packet spacing). This rate is then used to keep the bottleneck queue at a predetermined value.
• the routers are employing round-robin or some other fair service discipline.
• the approach does not work over heterogeneous networks, where the capacity of the reverse path could be orders of magnitude slower than the forward path because the data packet spacing is not preserved by the ACK packets.
• a receiver could employ a delayed ACK strategy as is common in many TCP implementations, and congestion on the reverse path can interfere with ACK spacing and invalidate the measurements made by the algorithm.
• congestion on the reverse path can interfere with ACK spacing and invalidate the measurements made by the algorithm.
• RTT round-trip time
• the proposed method uses variation in round-trip time (RTT) as the mechanism for determining congestion in the network.
• RTT monitoring alone cannot take into account the effects of congestion on the reverse path (as a contributing factor to increased RTT measurements).
• the present invention satisfies that need, as well as others, and overcomes deficiencies in current TCP-based techniques.
• the present invention is an extension of TCP that is capable of differentiating losses due to congestion from those caused by a lossy wireless link.
• the present invention is capable of determining congestion on the forward path of the connection and, therefore, can make a more accurate determination of the presence of congestion rather than a simple random loss on a lossy wireless channel.
• TCP-Santa Cruz can monitor the queue developing over a bottleneck link and thus determine whether congestion is increasing in the network. It can then identify losses as either congestive or random and respond appropriately. TCP-Santa Cruz also addresses the issue that applications over the Internet are likely to operate over paths which either exhibit a high degree of asymmetry or, which appear asymmetric due to significant load on the reverse data path.
• TCP-Santa Cruz is a new implementation of TCP that detects not only the initial stages of congestion in the network, but also identifies the direction of congestion; that is, TCP-Santa Cruz determines whether congestion is developing in the forward or reverse path of the connection. TCP-Santa Cruz is then able to isolate the forward throughput from events such as congestion that may occur on the reverse path.
• congestion is determined by calculating the relative delay that one packet experiences with respect to another as it traverses the network; this relative delay is the foundation of our congestion control algorithm.
• the relative delay is then used to estimate the number of packets residing in the bottleneck queue; the congestion control algorithm keeps the number of packets in the bottleneck queue at a minimum level by adjusting the TCP source's congestion window.
• the congestion window is reduced if the bottleneck queue length increases (in response to increasing congestion in the network) and the window is increased when the source detects additional bandwidth availability in the network (i.e., after a decrease in the bottleneck queue length).
• TCP-Santa Cruz can be implemented as a TCP option by utilizing the extra 40 bytes available in the options field of the TCP header.
• TCP-Santa Cruz can easily identify a congestive loss as one which is preceded by an increase in the bottleneck queue length.
• a wireless loss on the other hand, can be identified as a loss that is not preceded by a buildup in the bottleneck queue.
• Our protocol monitors changes in the bottleneck queue over an interval substantially equal to the amount of time it takes to transmit one window of data and receive acknowledgments corresponding to all the packets in the window. If a packet is lost during an interval preceded by two consecutive intervals in which the bottleneck queue increased, then we conclude that the loss was due to congestion in the network.
• FIG. 1 is a diagram illustrating an example of RTT ambiguity and determination of forward and reverse path queueing.
• FIG. 2 is a diagram illustrating the transmission of two packets and corresponding relative delay measurements.
• FIG. 3A through FIG. 3D are diagrams illustrating determination of forward path congestion for the packets transmitted as shown in FIG. 2 for the cases of equal delay, 1 st packet delayed, 2 nd packet delayed, and non-FIFO arrival of packets, respectively.
• FIG. 4 is a state diagram illustrating the determination of whether a Queue is filling, draining or maintaining based on the value of D ti .
• FIG. 5 is a diagram depicting a bottleneck link where delay consists of XQ which is the delay due to queueing and pktS, which is the packet service time over the link.
• FIG. 6 is a state diagram illustrating the congestion avoidance policy and slow start followed by the present invention.
• FIG. 7 is a state diagram illustrating monitoring buildup of network queue interval according to the present invention.
• FIG. 8 is a diagram showing a ACK window transmitted from a receiver to sender wherein packets #1 , #3 and #5 are lost.
• TCP-Santa Cruz can monitor the queue developing over a bottleneck link in a wireless network and thus determine whether congestion is increasing in the network. It can then identify losses as either congestive or random and respond appropriately.
• TCP-Santa Cruz can monitor the queue developing over a bottleneck link in a wireless network and thus determine whether congestion is increasing in the network. It can then identify losses as either congestive or random and respond appropriately.
• TCP-Santa Cruz comprises two major areas of improvement over TCP Reno: congestion control and error recovery.
• congestion control and error recovery we can perform timely and efficient early retransmission of any lost packet, eliminate unnecessary retransmissions for correctly received packets when multiple losses occur within a window of data, and provide RTT estimates during periods of congestion and retransmission, i.e., eliminate the need for Karn's algorithm.
• the congestion control algorithm allows us to determine when congestion exists in the forward path; a condition which cannot be detected by a round-trip time estimate. This type of monitoring allows us to detect the incipient stages of congestion and to respond to these early warning signs by increasing or decreasing the congestion window.
• FIG. 1 shows an example of the ambiguity involved when only RTT measurements are considered.
• the true cause of increased RTT for the second packet is congestion in the reverse path 14, not the data path 16.
• Our protocol solves this ambiguity by introducing the notion of the relative forward delay. Using this mechanism we will show that in the forward path congestion has actually improved for the second packet.
• the congestion control algorithm in TCP-Santa Cruz is based upon changes in the relative delay (increases and decreases in delay that packets experience with respect to each other) as packets are transmitted through the network.
• the relative delay is calculated from a timestamp returned by the receiver in every acknowledgment packet and specifies the arrival time of the packet at the destination.
• With relative delay measurements the sender can infer changes that are impossible to determine with a RTT measurements, i.e. the sender can determine whether congestion is increasing or decreasing in either the forward or reverse path of the connection. This determination can be made for every acknowledgment received (including retransmissions).
• FIG. 2 shows the transfer of two sequential packets, labeled pkt #1 (or / ) and pkt #2 (or/ ), transmitted from a source to a receiver.
• the sender maintains a table with the following times for each packet transmitted: the transmission time of the data packet; the arrival time of an acknowledgment packet for the packet; and the arrival time of the data packet at the destination as reported by the receiver in its acknowledgment. From this information the sender is able to calculate for any two data packets the following time intervals: S, the time interval between the transmission the packets; A, the time between arrival of their acknowledgments at the sender; and R, the inter-arrival time of the data packets at the receiver. From these values, the relative forward delay can be obtained:
• FIG. 3 illustrates how congestion is determined in the forward path.
• D ti 0 the two packets experience the same amount of queuing in the forward path.
• FIG. 3B shows that the first packet was delayed more than the second packet whenever D ti ⁇ 0.
• FIG. 3C shows that the second packet has been delayed with respect to the first one when D ti > 0.
• FIG. 3D illustrates out-of-order arrival at the receiver. In this case the sender is able to determine the presence of multiple paths to the destination by the timestamps returned from the receiver.
• FIG. 4 shows via a state diagram 20 how the computation of the relative forward delay, D h F i , provides information about the bottleneck queue and allows an easy determination of change in queue state based upon calculations of -Dr over time.
• the goal in TCP-Santa Cruz is to allow the network queues (specifically the bottleneck queue) to grow to a desired size; the specific algorithm to achieve this goal is described next.
• 1.2 Congestion Control Algorithm Summing the relative delay measurements over a period of data flow gives us an indication of the level of queuing at the bottleneck.
• the congestion control algorithm of TCP-Santa Cruz operates by summing the relative delays from the beginning of a session, and then updating the measurements at intervals equal to the amount of time to transmit a windowful of data and receive the corresponding ACKs.
• the relative delay sum is then translated into the equivalent number of packets (queued at the bottleneck) represented by the sum of relative delays. In other words, the algorithm attempts to maintain the following condition:
• BWDP bandwidth delay product
• the relative delay gives us a sense of the change in queuing, but it tells us nothing about the actual number of packets represented by this number.
• We translate the sum of relative delays into the equivalent number of queued packets by first calculating the average packet service time (sec/pkt) achieved by a session over an interval. This rate over the forward path is of course limited by the bottleneck link.
• Our model of the bottleneck link 30, depicted in FIG. 5, consists of two delay parameters: queuing delay, ⁇ Q; and output service time, pktS, the amount of time to service a packet.
• the queuing delay is variable and is controlled by the congestion control algorithm by changing the sender's congestion window. We get a sense of this value with the relative delay measurements.
• the output rate of a FIFO queue will vary according to the number of sessions and the burstiness of the arrivals from various sessions.
• the packet service rate is calculated as pktS - — , where R is the
• Adjustments are made to the congestion window only at discrete intervals, i.e., in the time taken to empty a windowful of data from the network. Over this interval, M ⁇ is calculated and at the end of the interval it is added to N, . If the result falls within the range of n + ⁇ , the congestion window is maintained at its current size. If, however, N,_ falls below n - ⁇ , the system is not being pushed enough and the window is increased linearly during the next interval. If N, ( rises above n + ⁇ , then the system is being pushed too high above our desired operating point, and the congestion window is decreased linearly during the next interval.
• the algorithm used by TCP at startup is the slow start algorithm used by TCP Reno with two modifications: the initial cwnd is set to two segments instead on one so that initial values for D tJ can be calculated and the algorithm may stop slow start before ssthresh > cwnd if any relative delay measurement or N,, exceeds n/2.
• the congestion window doubles every round-trip time, leading to an exponential growth in the congestion window.
• Many divergent problems have been identified with slow start.
• the algorithm actually causes the congestion window to grow very quickly and such rapid growth often leads to congestion in the data path. This is the approach taken by TCP Reno to find the optimal window size.
• slow start is actually too slow, and if a transfer is very short it is possible that the pipe is never filled, which leads to underutilization of the network and renders short transfers very expensive.
• TCP-Santa Cruz provides an improved RTT estimate over traditional TCP approaches, as well as TCP Vegas, by measuring the round-trip time (RTT) of every segment transmitted. RTT estimates are made for every packet for which an acknowledgment (ACK) is received, including retransmissions. This eliminates the need for Karn's algorithm (in which RTT measurements are not made for retransmissions) and timer-backoff strategies (in which the timeout value is essentially doubled after every timeout and retransmission).
• TCP-Santa Cruz requires each returning acknowledgment packet to indicate the precise packet that caused the ACK to be generated and the sender must keep a timestamp, for each transmitted packet.
• the precise packet can be uniquely identified by specifying both a sequence number and a transmission copy number. For example, the first transmission of packet "1" is specified as "1.1”, the second transmission is labeled "1.2”, etc. In this way, it is possible for the sender to perform a new round-trip time estimate for every acknowledgment it receives.
• acknowledgments from the receiver are logically a 3-tuple consisting of a cumulative acknowledgment (indicating the sequence number of the highest in-order packet received so fax), and the 2-element sequence number of the packet generating the ACK (usually the most recently received packet). For example, acknowledgment "5.7.2” indicates a cumulative acknowledgment of "5" and the current acknowledgment was generated by the second transmission of a packet with sequence number "7".
• TCP-Santa Cruz As congestion develops in a wired network, data packets fill the network links. If the rate of data entering the network continues to increase, data packets fill the network queues until the queues reach capacity and the routers begin to drop packets. In other words, losses due to congestion are preceded by an increase in the network bottleneck queue.
• the basic methods used to detect congestion in TCP-Santa Cruz are easily applied to networks containing wireless links and are ideal for distinguishing congestion versus random loss in a network which may or may not contain a wireless link.
• TCP-Santa Cruz easily identifies a congestive loss as one which is preceded by an increase in the bottleneck queue length.
• a wireless loss on the other hand, can be identified as a random loss that is not preceded by a buildup in the bottleneck queue.
• TCP-Santa Cruz monitors changes in the bottleneck queue over an interval equal to the amount of time it takes to transmit one window of data and receive acknowledgements corresponding to all the packets transmitted in the window.
• FIG. 7 is a state diagram 50 that shows how the protocol counts intervals of queue buildup.
• TCP-Santa Cruz reduces the transmission rate only when congestion is identified as the cause of lost packets; otherwise, wireless losses can simply be retransmitted without a reduction in the data transmission rate.
• the receiver in TCP-SC returns an ACK Window to the sender to indicate any holes in the received sequential stream.
• the ACK Window is similar to the bit vectors used in previous protocols, such as NETBLT and TCP-SACK. However, unlike TCP-SACK our approach provides a new mechanism whereby the receiver is able to report the status of every packet within the current transmission window. Note that TCP-SACK is generally limited by the TCP options field to reporting only three sections of continuous data within a window.
• the ACK Window is maintained as a vector in which each bit represents the receipt of a specified number of bytes beyond the cumulative acknowledgment.
• the receiver determines an optimal granularity for bits in the vector and indicates this value to the sender via a one-byte field in the header.
• a maximum of 19 bytes are available for the ACK window in order to meet the 40-byte limit of the TCP option field in the TCP header.
• the granularity of the bits in the window is bounded by the receiver's advertised window and the 19 bytes available for the ACK window; this can accommodate a 64K window with each bit representing 450 bytes.
• a bit in the vector would represent the MSS of the connection, or the typical packet size. Note this approach is meant for data intensive traffic, therefore bits represent at least 50 bytes of data. If there are no holes in the expected sequential stream at the receiver, then the ACK window is not generated.
• FIG. 8 shows the transmission of five packets, #1 , #2, #3, #4, #5, three of which (#1 , #3, and #5) are lost and are shown using a stipple pattern.
• the packets are of variable size and the length of each is indicated by a horizontal arrow.
• Each bit in the ACK window represents 50 bytes with a "1” if the bytes are present at the receiver and a "0" if they are missing.
• the sender knows that this range includes packets "3” and "4" and is able to infer that packet "3” is lost and packet "4" has been received correctly.
• the sender maintains the information returned in the ACK Window, flushing it only when the window advances. This helps to prevent the unnecessary retransmission of correctly received packets following a timeout when the session enters slow start.
• Non-trigger is when a packet is retransmitted by the sender without previous attempts, i.e., TCP's fast retransmission mechanism never happened. For this reason we need to quickly recover losses without necessarily waiting for three duplicate acknowledgments from the receiver.
• TCP-Santa Cruz Given that TCP-Santa Cruz has a much tighter estimate of the RTT time per packet and that the TCP-Santa Cruz sender receives precise information on each packet correctly received, TCP-Santa Cruz is able to determine when a packet has been dropped and can avoid waiting for TCP's Fast Retransmit algorithm and quickly retransmit and recover a lost packet once any ACK for a subsequently transmitted packet is properly received and a time constraint is met.
• packet / ' initially transmitted at time ti, is lost and is marked as a hole in the ACK window.
• Packet / can be retransmitted as soon as an acknowledgment arrives for any packet transmitted at time t x , such that t x > t,, and tc ⁇ i ⁇ ent - ti > RTT e , where tcurrent is the current time and RTT e is the estimated round-trip time of the connection. Any packet which is marked as unreceived in the ACK window can be a candidate for early retransmission. 4.
• TCP-Santa Cruz can be implemented as a TCP option containing the fields depicted in Table 1.
• the TCP-SC option can vary in size from 1 1 bytes to 40 bytes, depending on the size of the acknowledgment window (see Section 3.1 ). 5. Performance Results
• FIG. 7 illustrates how TCP-Santa Cruz counts the consecutive intervals over which it notices a bottleneck queue increases in order to determine if congestion is present when a loss is detected.
• the value of count was rarely equal to two, which would indicate that nearly all losses on the wireless link would be considered as random by the protocol. In other words, once the losses are discovered, we expect the protocol to simply retransmit most losses without reducing the transmission window.
• n 4
• TCP-Santa Cruz would try to fill the bit pipe and maintain an extra 4 packets in the bottleneck queue (at the base station).
• TCP-Santa Cruz maintained its window around 10 packets, exactly the BWDP (6 packets) plus the extra 4 packets in the queue.
• This steady value of the congestion window is directly correlated to the number of packets in the bottleneck queue.
• the congestion window for TCP Reno oscillated wildly between 1 and 25 packets (ant at times up to 60 packets during the Fast Recovery phase). This variation in the congestion window was directly responsible for long end-to-end delays experienced by transfers using TCP Reno.
• TCP Reno did not experience an appreciable reduction in throughput because it was able to recover most errors via the fast transmission mechanism and therefore was able to keep its window large enough to fill the bit pipe of the connection.
• TCP Reno experienced a very large delay variance at the rate 1/12.5 packet loss. This was because packets either make it though quickly with a small bottleneck queue, or they experience a long delay due to timeouts. The delay in TCP-Santa Cruz, however, was fairly steady until the large error rate at which time so many dropped packets were also dropped on the retransmission that timeouts were experienced as well.
• TCP-Santa Cruz is able to maintain high throughput over the wireless link because it does not reduce its sending rate when the losses are determined to be due to the wireless link.
• TCP Reno on the other hand, can make no such distinction and as a result performs very poorly, even when link error rates are low.
• TCP-Santa Cruz also produces reduced delay and delay variation from source to receiver because it keeps the bottleneck queue at a minimum level and does not create the oscillations in bottleneck queue length that is typical of the TCP Reno transmission pattern. Additional details regarding TCP-Santa Cruz and its performance can be found in C. Parsa and J.J.
• TCP-Santa Cruz makes use of a simple timestamp returned from the receiver to estimate the level of queuing in the bottleneck link of a connection.
• the protocol successfully isolates the forward throughput of the connection from events on the reverse link by considering the changes in delay along the forward link only. We successfully decouple the growth of the congestion window from the number of returned ACKs (the approach taken by TCP), which makes the protocol resilient to ACK loss.
• the protocol provides quick and efficient error-recovery by identifying losses via an ACK window without waiting for three duplicate acknowledgments. An RTT estimate for every packet transmitted (including retransmissions) allows the protocol to recover from lost retransmissions without using timer-backoff strategies.
• TCP-Santa Cruz is the only protocol to date that can accurately determine random wireless losses by monitoring the increase in queue length along the forward path of a connection. In this way, there is no chance for fluctuations in congestion along the reverse path in influence our determination of congestion along the forward path.
• TCP-Santa Cruz provides high throughput and low end-to-end delay and delay variance over networks with a simple bottleneck link, networks with congestion in the reverse path of the connection, and networks which exhibit path asymmetry.
• TCP-Santa Cruz eliminates the oscillations in the congestion window, but still maintains high link utilization. As a result, it provides much lower delays than current TCP implementations.
• our protocol provides a 20% - 39% improvement in end-to-end delay (depending on the value of n) and a delay variance three orders of magnitude lower than TCP Reno.
• TCP-Santa Cruz provides an improvement in throughput of at least 47% - 67% over both TCP Reno and TCP Vegas, as well as an improvement in end-to-end delay of 45% to 59% over TCP Reno with a reduction in delay variance of three orders of magnitude.
• TCP Reno and TCP Vegas achieve link utilization of only 52% and 33%, respectively, whereas TCP-Santa Cruz achieves 99% utilization. End-to-end delays for this configuration are also reduced by 42% - 58% over TCP Reno.

Priority Applications (1)

Applications Claiming Priority (2)

Publications (2)

Family

ID=22862387

Family Applications (1)

Country Status (3)

Cited By (8)

Families Citing this family (97)

Family Cites Families (16)

• 2001
  • 2001-08-30 WO PCT/US2001/027150 patent/WO2002019654A2/en active Application Filing
  • 2001-08-30 US US09/945,106 patent/US7200111B2/en active Active
  • 2001-08-30 AU AU2001288589A patent/AU2001288589A1/en not_active Abandoned
  • 2001-10-29 US US10/021,318 patent/US7020083B2/en not_active Expired - Lifetime

Non-Patent Citations (4)

Also Published As

Similar Documents

Legal Events