(19) United States
`(12) Patent Application Publication (10) Pub. No.: US 2002/0186660 A1
`Bahadiroglu
`(43) Pub. Date:
`Dec. 12, 2002
`
`US 2002O186660A1
`
`(54)
`
`(76)
`
`(21)
`(22)
`
`(51)
`(52)
`
`ADAPTIVE CONTROL OF DATA PACKET
`SIZE IN NETWORKS
`
`Inventor: Murat I. Bahadiroglu, Bedford, NH
`(US)
`Correspondence Address:
`DAVIS & BUJOLD, P.L.L.C.
`500 NORTH COMMERCIAL STREET
`FOURTH FLOOR
`MANCHESTER, NH 03101 (US)
`Appl. No.:
`09/879,761
`
`Filed:
`
`Jun. 12, 2001
`
`Publication Classification
`
`Int. Cl. ................................................... H04J 3/22
`U.S. Cl. ............................................ 370/248; 370/468
`
`ABSTRACT
`(57)
`An adaptive packet mechanism and method for optimizing
`data packet transmission through a network connection
`between a Sending node and a receiving node. Current
`network conditions in the connection are periodically deter
`mined wherein the network conditions pertain to the latency
`and jitter of packet transmission between the Sending node
`and receiving node. The measurements of latency and jitter
`are used to determine an optimum packet Size and an
`optimum inter-packet interval for transmission of packet
`data between the Sending node and the receiving node and
`are used in the transmission of data packets from the Sending
`node to the receiving node. Network conditions may be
`determined by transmission of monitor or data packets and
`may be determined at either or both of the sending or
`receiving nodes and the optimum packet Size and inter
`packet interval are determined by a fuzzy logic analyzer, a
`neural network analyzer or a combined fuzzy logic/neural
`network analyzer.
`
`
`
`NODE 12S
`
`NODE 12R
`RESULT
`DATABASE
`CONDITION
`RECORD
`
`ENTRY 4OE
`
`16M
`16MT, 16MS,
`16MN, 16MP
`
`DATA
`RECIPIENT
`
`Samsung Ex. 1006
`Page 1
`
`

`

`Patent Application Publication Dec. 12, 2002 Sheet 1 of 10
`
`US 2002/0186660 A1
`
`OLTSSOT??×oe) –
`
`
`
`Samsung Ex. 1006
`Page 2
`
`

`

`Patent Application Publication Dec. 12, 2002 Sheet 2 of 10
`
`US 2002/0186660 A1
`
`M
`
`M
`
`M
`
`m
`
`- d
`Y4
`O
`(5
`n
`
`
`
`()
`O
`
`5
`N4
`CD
`(V
`a
`
`:
`
`OeASSO exoe
`
`S
`o
`
`O
`O CD
`O N
`Y
`of
`s
`s
`S Vs
`O n
`
`O
`
`:
`
`Samsung Ex. 1006
`Page 3
`
`

`

`Patent Application Publication Dec. 12, 2002 Sheet 3 of 10
`
`US 2002/0186660 A1
`
`
`
`009||
`
`009
`
`e SSO exoe
`
`Samsung Ex. 1006
`Page 4
`
`

`

`Patent Application Publication Dec. 12, 2002 Sheet 4 of 10
`
`US 2002/0186660 A1
`
`
`
`Samsung Ex. 1006
`Page 5
`
`

`

`Patent Application Publication Dec. 12, 2002 Sheet 5 of 10
`
`US 2002/0186660A1
`
`
`
`N
`
`t
`O
`Ol
`C
`
`uOISS?S
`
`e º
`
`S.
`2
`
`øø º ?ºpOWN ISO
`
`
`
`
`
`
`
`
`
`
`
`Samsung Ex. 1006
`Page 6
`
`

`

`Patent Application Publication Dec. 12, 2002 Sheet 6 of 10
`
`US 2002/0186660 A1
`
`
`
`| alia || aos | avis
`
`}J?Áë?
`
`
`
`d. LNN
`
`
`
`
`
`
`
`Samsung Ex. 1006
`Page 7
`
`

`

`Patent Application Publication Dec. 12, 2002 Sheet 7 of 10
`
`US 2002/0186660 A1
`
`
`
`Samsung Ex. 1006
`Page 8
`
`

`

`Patent Application Publication
`
`
`
`
`
`Samsung Ex. 1006
`Page 9
`
`

`

`Patent Application Publication Dec. 12, 2002. Sheet 9 of 10
`
`US 2002/0186660 A1
`
`Test Schedule Script File 38
`
`Header Name
`
`TestType
`
`Net Condition
`
`Data Transfer
`
`IP Protocol
`
`UDP
`
`Send IP Address
`
`-1
`
`
`
`Send Embed
`F
`Recv Mode
`
`
`
`Recy
`
`Rec IP Address
`555
`Recy IP Port
`R 5552
`Recw Embed
`Offset Sec
`Force Payload
`Size
`
`28
`10-400
`
`
`
`
`
`10-100
`
`Force Packet
`Count
`Inter-packet
`arbitra
`Delay ms
`Loop and Restart Net Condition
`Data Transfer
`
`
`
`
`
`
`
`Description
`|
`Records tests of 1028 bytes packet transmission at a
`specified inter-packet delay.
`Records test of packets with a forced payload size or
`forced packet count at a specified inter-packet delay,
`UDP protocol provides minimal transmission delay
`because it excludes the connection setup process, flow
`control and retransmission - no error checking
`performance.
`
`
`
`T Sender's IP address
`Negative number indicates IP stack chooses the
`sending port
`
`Padding
`Net Condition receive mode - test always functions in
`a loop
`
`Data Transfer receive mode -- data is transmitted one
`WaW.
`Receiver's IP address
`Net Condition test port ID
`Data Transfer test port ID
`Padding
`
`Net Condition test packet sizes
`Data Transfer test packet sizes
`
`Data Transfer test packet breakdown
`Inter-packet delay for Net Condition and Data
`Transfer tests are determined by TSSF.
`Yes - data transfer is constantly repeated
`No - data transfer is not repeated
`
`FIG. 6B
`
`
`
`
`
`
`
`
`
`
`
`Samsung Ex. 1006
`Page 10
`
`

`

`Patent Application Publication Dec. 12, 2002 Sheet 10 of 10
`
`US 2002/0186660 A1
`
`NETWORK CONDITION RECORD 40R
`
`44A - packet number - the number of packets that were sent for the whole data
`44B - payload - the size of a sent packet in bytes
`44C - total bytes - the total aggregate bytes of the combined payloads that represent the total data
`44D - Interpacketdelay ms - the time spacing of packets sent (in ms)
`44E - elapsedtime sec - the duration the listener ran
`44F - packetsIN - the # of packets received IN by listener per TRANSFER
`44G - packetsOUT - the # of packets sent OUT by listener per TRANSFER
`44H - bytes N - the bytes received in per transfer
`44
`bytesOUT - the bytes sent out per transfer
`44G - bytes.IN/sec - the rate of bytes
`44H - bytes.OUT/sec - the rate of bytes OUT
`44 - packetsIN/sec - the rate of packects IN
`44J - packetsOUT/sec - the rate of packets OUT
`44K - numreads - the number of attempted reads from socket
`44L - nunreadsblocked - the number of reads to socket having nothing to read
`44M - numkrnlreaderrs - the number of read errors (e.g. MAC layer failure,
`helpful to show wireless network blackout conditions)
`44N - numwrites - the number of attempted writes to socket
`44O - numwritesblocked - the number of writes to socket unable to send out (shows
`that network is clogged; IP stack unable to accept more)
`44P - numkernelwriteerros - the number of write errors (see note on MAC layer)
`44Q - numkernel unspecifiederrors - the number of unknown errors
`44R - jitter avg sec - the avgjitter
`44S- jitter max sec - the max jitter
`44T - jitter min sec - the minjitter
`44U - packet loss - packets not received (the packet loss)
`44V - % loss - percentage of loss out of total
`44W - pass sequence - # of packets in sequence (by checking the incremental
`sequence ID on each packet. Next packet ID is current
`ID + 1)
`44X - fail sequence - # of packets not in precise sequence
`44Y- % out of sequence - percentage of packets out of sequence out of total
`net condition 2-way delay average - the two-way, round trip packet delay
`as measured in seconds, to microsecond precision. The average value across all packets received
`during the specific data transfer and its time frame.
`44Z - net condition 2-way delay max - the max of the delay value for the network condition
`44AA-net condition 2-way delay min - the min of the delay value for the network condition
`44AB-net condition jitter avg sec - the average jitter for the network condition
`44AC-trendupordown - the change of the delay during the test
`(+ means the delay was increasing during the test, - means
`it was decreasing
`
`Samsung Ex. 1006
`Page 11
`
`

`

`US 2002/0186660 A1
`
`Dec. 12, 2002
`
`ADAPTIVE CONTROL OF DATA PACKET SIZE IN
`NETWORKS
`
`FIELD OF THE INVENTION
`0001. The present invention is directed to a method and
`System for transmission of data through a network and, in
`particular, to a method and System for the transmission of
`data through a network wherein data packet Size is adapted
`to network conditions to enhance the data transmission rate
`to meet predetermined requirements of data transmission.
`
`BACKGROUND OF THE INVENTION
`0002 Networks have long been commonly used to com
`municate information of all forms between various computer
`Systems, Such as program code and data, including audio,
`graphic and Video or graphic information. As a result,
`computer Systems and communications Systems have
`evolved together, each influencing the other and each evolv
`ing to address methods and problems presented by the other.
`0.003
`For example, telephone networks, long being in
`existence as a widely available means of communicating
`information from one point to another when computers were
`first developed, were one of the first forms of network used
`for communication between computer Systems. The inherent
`characteristics of the original telephone networks thereby, as
`a consequence, defined the original computer communica
`tion methods and technologies. AS is well known and
`understood, telephone networks allow communication of
`audio data, or more broadly audio signals, between two or
`more users by establishing a dedicated communication cir
`cuit or “channel” between the users through one or more
`central Switching Systems. AS is also well known and
`understood, information in a computer System is Structured
`as units or packets of multiple bits or bytes of predetermined
`Size and, as a consequence, various methods and Signal
`formats were developed to communicate digital information
`packets through the telephone networkS.
`0004. The essential nature of the original telephone net
`WorkS is Such that a channel, once established, is dedicated
`exclusively to a single information exchange, whether the
`information is a conversation between user's or digital
`information being communicated between one computer
`and another. Dedicated channels are advantageous in that the
`entire bandwidth of the channel is available for a given
`information eXchange and in that any transmission delayS
`from an information Source to an information recipient are
`purely a function of the direct transmission Speed through
`the channel. Since dedicated channel transmission Speed
`does not significantly vary over time or with the information
`transmission load, dedicated channels are capable of pro
`Viding "isochronous transmission. A significant disadvan
`tage of dedicated channels, however, is that dedicated chan
`nels preempt Significant bandwidth of the communications
`network. That is, the complete bandwidth of a channel is and
`remains available and dedicated to a given link between
`computer Systems, even when the information exchange
`does not require the full bandwidth of the channel and even
`when no information is being transmitted.
`0005 For the above reasons, dedicated channels have
`been inadequate to meet the needs of intercomputer com
`munication as information exchange between computer Sys
`tems has become more common and widespread. AS a
`
`consequence, telephone Systems and computer communica
`tion methods have evolved to meet to the increased demand
`for information eXchange between computers, and have
`generally evolved in conjunction to meet common needs, in
`particular the transmission of data, and to make advanta
`geous use of jointly developing technologies.
`0006 For example, various forms of networks have been
`developed that are dedicated to or Specifically adapted to
`communication between computers. NetworkS Specifically
`oriented to intercomputer communications typically provide
`individual computers shared access to a common commu
`nication backbone having relatively broadbandwidth, Such
`as fiber optic cables or coaxial or twisted pair cable. The
`individual networked computers, commonly referred to as
`"nodes', are typically granted access to the complete band
`width of the backbone for the transmission of each infor
`mation packet for a period, and when a transmitting com
`puter completes transmission of a packet the backbone is
`made immediately available for the other computers or
`nodes connected to the network. Each computer thereby has
`access to the full bandwidth of the network, but only during
`the actual transmission of a packet, So that many computers
`may efficiently share the common “backbone'. A typical
`example of Such a computer network is a personal computer
`(“PC”) network as PCs are used for a wide range of
`applications requiring the PCs to communicate with each
`other over a computer network. Typical computer networks
`connecting PCs, as well as other computers, nodes and
`related devices and Systems, include local area networks
`(“LANs”) interconnecting PCs located in relatively close
`proximity and wide area networks (“WANs”), which are
`typically comprised of a network of Separate LANS and
`provide communications over wider areas. The “Web”, for
`example, is a very extensive WAN interconnecting LANS as
`well as PCs, larger computers, nodes and related devices.
`0007 Telephone networks in various forms, however,
`remain the most common method for interconnection and
`communication between computers. For example, telephone
`networks are frequently employed as a WANS linking indi
`vidual computers or LANs and the World Wide Web is
`primarily based upon telephone networkS. Many telephone
`Systems are presently implemented with broadband chan
`nels, Such as fiber optic lines, to provide increased numbers
`of channels or increased channel bandwidths for the trans
`mission of computer data and digitized Voice Signals. AS a
`consequence, there have been extensive efforts to establish
`interface Standards, methods and technologies for the faster
`and more efficient transmission of digital information pack
`ets through the various implementations of telephone net
`WorkS. For example, many telephone Systems are now
`implemented using computer network technologies, Such as
`broadband backbones, and many telephone networks also
`digitize audio information, that is, Voice Signals, into digital
`data packets analogous to computer data packets. The adap
`tation of computer network technologies to telephone net
`Works thereby allows Voice information to be communicated
`in a manner analogous to computer information packets by
`concurrent communication of multiple channels on a Single
`line and the routing of digitized voice packets through
`multiple paths. The adaptation of computer originated tech
`nology and methods to telephone Systems and Voice com
`munications thereby allow telephone networks to carry both
`Voice information and computer information with equal
`facility. The consequences of Such developments in tele
`
`Samsung Ex. 1006
`Page 12
`
`

`

`US 2002/0186660 A1
`
`Dec. 12, 2002
`
`phone Systems may be seen, for example, in the implemen
`tation and rapid expansion of Work Wide Web communica
`tions between computers, which is primarily implemented
`through the telephone networkS.
`0008. The development of the hardware technologies of
`telephone networks and of computer Specific networks has
`been paralleled by the development of Standards and proto
`cols for the more efficient transmission of data packets
`through computer networks. In this regard, the heart of any
`computer network, whether based upon a telephone network
`or implemented as a network dedicated to or Specifically
`adapted to communication between computers, is a commu
`nication protocol wherein a communications protocol is a Set
`of conventions or rules that govern the transfer of data
`between computer devices. The simplest protocols define
`only a hardware configuration, while more complex proto
`cols define timing, data formats, error detection and correc
`tion techniques and Software structures. Virtually all forms
`of computer networks commonly employ multiple layers of
`protocols wherein each layer of the protocols defines and
`controls a corresponding level of operations in the commu
`nication between computers through the network. For
`example, the physical protocol layer typically define and
`control the lowest level of operations and will commonly
`include a low-level physical or device layer protocol that
`assures the transmission and reception of a data Stream
`between two devices, while a higher level data link protocol
`layer defines and manages the construction and transmission
`of data packets. Network and transport layer protocols, Such
`as the Transmission Control Protocol/Internet Protocol
`(“TCP/IP”), operate a higher levels than the physical layer
`protocols, and in turn govern the transmission of data
`through the network itself, thereby ensuring the reliable
`end-to end delivery of the data packets defined at the lower
`levels.
`0009. There are presently a range of industry standard
`protocols for the efficient transmission of data through both
`computer Specific networks and telephone networks, most of
`which are based upon the packet transmission of data.
`Examples of existing Standards and protocols for the trans
`mission of computer data through networks, including tele
`phone networks and computer Specific networks, include the
`Integrated Services Digital Network (“ISDN”), which stan
`dardizes connection interfaces, transmission protocols and
`Services to create a unified digital circuit-Switching network.
`A further example is Broadband ISDN (“BISDN”) which,
`unlike ISDN which is a digital network standard, uses packet
`relay, or Asynchronous Transfer Mode ("ATM"), as a trans
`mission Standard, particularly for transmission over broad
`band “backbones”. ATM is primarily a connection-oriented
`technique that can transport both connection and connec
`tionless-oriented Services at either a constant bit rate or a
`variable bit rate. ATM provides bandwidth on demand and
`handles all traffic types through fast-packet Switching tech
`niques that reduce the processing of protocols and uses
`Statistical multiplexing to control transmission through the
`network.
`0010. It has long been commonly recognized and under
`stood, however, that a primary limitation upon the commu
`nication of data through networks is that the bandwidth
`resources available in any network are limited. Other, related
`problems include network latency and jitter, which are
`delays in end-to-end data transmissions arising from a
`
`variety of causes. The “latency” of a network includes the
`inherent delay times through the elements of the network,
`Such as through the cables or lines, the Switching routers and
`the communications processes or protocols being executed
`at each end of the connection. Although the latency of a
`network tends to be relatively constant for a given connec
`tion, latency may vary from connection to connection,
`depending upon the particular route taken through a network
`by a given connection and upon the traffic load of the
`network or in the route defined for the connection. The
`latency of a network effects the overall efficiency and
`effective bandwidth or capacity of a network by imposing a
`practical limit on the rate at which data, and in particular
`packet data, can be communicated through the network,
`particularly Since many protocols require an acknowledg
`ment of each packet before a next packet is transmitted.
`"Jitter', in turn, may be defined as change in network latency
`as a function of time. Jitter is typically unpredictable and
`may arise from a number of causes, but most frequently
`results from variations in the traffic load along the route
`taken by a connection through a network. AS Such, the
`"jitter of a network connection may vary Significantly from
`packet to packet and even more So in Systems that allow
`independent routing of individual packets. Jitter also effects
`the overall efficiency and effective bandwidth or capacity of
`a network by imposing a practical limit on the rate at which
`data can be communicated through the network, and has
`particularly adverse effects in asynchronous networks, Such
`as many computer networks and networks employing, for
`example, the ATM protocol.
`0011. It has long been understood that efficient bandwidth
`and traffic management is essential to obtain full advantage
`of the bandwidth and capacity of a network and Significant
`efforts have been made in designing traffic flow and con
`gestion control processes, bandwidth management mecha
`nisms and routing algorithms to manage available network
`bandwidth and capacity. The goal of Such developments has
`been a network that is able to transmit a useful level of traffic
`that is directly proportional to the traffic offered to the
`network up to the maximum transmission capacity of the
`network, and thereafter to continue to operate at the maxi
`mum network capacity regardless of the imposed traffic
`load. The actual performance achieved in networks falls far
`Short of these goals, however, for reasons pertaining to
`practical constraints in implementing networks and arising
`from limitations inherent in the methods presently used to
`manage networkS.
`0012 For example, the simplest method for obtaining
`Satisfactory performance in a network is to oversize the
`equipment, that is, the number and capacity of the channels
`and the traffic capacity of the routing Switches, So that in all
`anticipated operating conditions the network will be oper
`ating in a Zone that is well distant from congested levels or
`Zones of operation. This method, however, is generally
`unsatisfactory because of cost and because the traffic
`demands on networks historically increaseS rapidly over
`time to exceed the anticipated maximum loads on the
`networks, thus resulting in eventual congestion regardless of
`the initial capacity of a network.
`0013 The preferred methods for obtaining satisfactory
`performance in networks have thereby focused on traffic
`management methods for managing and controlling the
`traffic load and flow in a network, mechanisms for allocating
`
`Samsung Ex. 1006
`Page 13
`
`

`

`US 2002/0186660 A1
`
`Dec. 12, 2002
`
`network bandwidth and improvements in protocols for the
`efficient transmission of data. Traffic flow and control mea
`Sures, for example, typically include flow control for regu
`lating the packet transmission rate of a transmitting System
`to a rate compatible with the rate the receiving System can
`absorb. Load regulation mechanisms, in turn, globally limit
`the number of packets present at any time in the network to
`thereby avoid overloading the capacity of the channels and
`routing Switches while load balancing mechanisms distrib
`ute traffic over the links of the network to avoid local
`congestion.
`0.014) A traffic management system, bandwidth allocation
`mechanism or communications protocol, however, must be
`Structured to meet requirements arising from the character
`istics of network traffic which are difficult to accommodate.
`For example, network traffic is typically not well behaved,
`that is, the traffic load may vary widely and at unpredictable
`times or under unpredictable circumstances, and often
`departs Substantially from the initially assumed traffic
`parameters. If Such a departure persists, the traffic manage
`ment System must, for example, assign one or more new
`connection bandwidths to one or more connections to
`accommodate the new traffic parameters, which in turn may
`affect yet other connections and require adaptations or
`modifications of those connections, and So on. Determining
`the appropriate responses of the traffic management System
`to radical changes in traffic behavior in turn presents yet
`further problems. Typical problems are those of obtaining
`and filtering traffic load measurements to separate transient
`changes of traffic behavior from longer term changes and
`determining ranges within which the initially assumed traffic
`parameters can be maintained and outside of which new
`connection bandwidths must be requested. For example, a
`bandwidth that is too large for the actual traffic is wasteful
`of connection resources while a bandwidth that is too small
`results in excessive packet loSS. Continuous adjustments of
`channel bandwidths to meet transient conditions, however,
`absorb excessive network resources in continuous adapta
`tion to conditions that may have changed by the time the
`adaptations take effect.
`0.015. Another source of network problems arises when
`the communications protocols implemented in a given net
`work for efficient data transmission conflict with the opera
`tion of the network traffic management System. This prob
`lem arises, for example, from conflicting goals when the
`traffic manager operates to optimize the overall performance
`of the network while the protocols executed in the network
`users attempt to optimize the performance of each user of the
`network individually. For example, it has been described
`above that TCP/IP is a very commonly used protocol.
`TCP/IP, however, employs a “go back N method” for
`dealing with errors and flow control problems over a net
`work. If there is a transmission error, a packet loss, excessive
`latency in the delivery of a packet, delivery of a packet out
`of Sequence or an overflow of a receiver buffer, the protocol
`retransmits N preceding packets. The retransmission of
`packets by TCP/IP in response to problems that may arise
`from traffic congestion may thereby Significantly increase
`the traffic load congestion that the network management
`System is attempting to address. This problem may become
`particularly severe if the TCP/IP protocols of several users
`are retransmitting packets in response to data communica
`
`tions problems arising from traffic congestion rather than
`from circumstances in the individual transmitters or recipi
`entS.
`0016 To illustrate the above by considering various
`mechanisms of the protocols and network management
`mechanisms of the prior art in greater detail, many protocols
`such as TCP/IP implement packet priority mechanisms to
`prevent Saturation of a network. A "leaky bucket' algorithm,
`for example, limits the number of low priority packets that
`can be transmitted in a fixed period of time when the packets
`digreSS from an original rate of transmission, So that high
`priority traffic is thereby transmitted with little or no delay.
`A leaky bucket or similar packet priority mechanism will
`thereby maintain packet transmission at acceptable levels if
`the traffic is not unreasonable and if the traffic remains
`within a reasonable range of the initial transmission param
`eters. Network traffic loads, however, are often unreasonable
`and often depart Substantially from the initial transmission
`parameters, thereby requiring other mechanisms to accom
`modate changes in the traffic loads.
`0017 For example, if a departure from the initial traffic
`parameters in a given connection persists for a significant
`length of time, a traffic control System will typically assign
`a new connection bandwidth to the connection to accom
`modate the new traffic parameters. This mechanism, how
`ever, must adapt both to radical changes in traffic behavior
`and the problem of short-lived changes in traffic behavior as
`opposed to long term changes and must determine which
`connection bandwidths would be suitable to traffic behavior
`at a given time. Too large a bandwidth would waste con
`nection resources and a too small a bandwidth would result
`in packet loSS.
`0018. The problem of determining appropriate connec
`tion bandwidths may be illustrated by consideration of the
`problems of latency, jitter and "burst' type data. That is, and
`for example, if packets are being transmitted through a
`connection with no losses and the predefined maximum
`window Size is appropriate, wherein window Size may be
`considered as connection bandwidth and the time interval
`assigned for transmission to a given System, the Steady flow
`of transmission data will bring the TCP protocol to a steady
`State. In this steady State, one new packet of data is placed
`on the network for transmission each time an acknowledg
`ment of receipt of the previous packet is received from the
`receiving end node of the connection by the Sending end
`node of the connection. The time lapse between each packet
`transmission is thereby determined by the rate at which
`acknowledgments arrive at the Sending end node. If network
`bandwidth and latency remain consistent, packets flow
`freely and few packets are lost to buffer overflows, such as
`in routers. However, and as discussed above, many networks
`are Subject to long term changes in connection bandwidth, as
`when bandwidth is reassigned due to changing traffic con
`ditions, and long and Short term variations in latency,
`including jitter due, for example, to Short term traffic con
`ditions. Variations in bandwidth requirements may also arise
`from the nature of the data being transmitted. For example,
`Video and Voice data typically appears as data “clusters'
`interspersed with large gaps. As a consequence, the Voice or
`Video data clusters are typically transmitted in "bursts' and
`as a result a protocol such as TCP/IP does not make efficient
`use of the bandwidth available between bursts.
`
`Samsung Ex. 1006
`Page 14
`
`

`

`US 2002/0186660 A1
`
`Dec. 12, 2002
`
`0019. A traffic control system typically adapts to changes
`in traffic parameters by defining the regions of the network
`where bandwidth allocations do and do not require adjust
`ment to meet traffic needs. Bandwidth allocations are
`adjusted upward if measurements indicate the anticipated
`maximum acceptable packet loSS will be exceeded or that
`traffic on the connection will interfere with other connec
`tions sharing the transmission medium. Bandwidth require
`ment is adjusted downward if the quality of service will be
`acceptable for all users.
`0020. To determine the required bandwidth adaptations, a
`traffic control System will take measurements of the mean
`burst duration and mean bit rates and will filter the mea
`Surements to insure that a Statistically reliable number of raw
`measurements have been obtained. The minimum required
`number of raw measurements and the mean bit rate of the
`traffic will determine the time required to collect the raw
`measurements. This measurement time, in turn, may used to
`analyze Statistics of the incoming data Stream to, for
`example, a leaky bucket mechanism and to determine the
`effect of the leaky bucket on incoming traffic. This effect
`may then be used to predict how the leaky bucket is
`monitoring the transmission, and to determine packet loSS
`probability.
`0021 When traffic parameters fall outside of the accept
`able bandwidth adaptation range, a traffic control System
`will request a new connection with a different bandwidth.
`Typically, however, the adaptation mechanism requires a
`longer time to adapt to upward changes in the traffic param
`eters, that is, to increases in traffic load or Speed, due to the
`time lag in the measurement and filtration process. This, in
`turn, often results in overcompensation in upward or down
`ward network bandwidth adjustments and either inefficient
`use of network resources or degradation of data transmission
`or both. This problem further compounded in that it is
`common to have a single processor monitoring and perform
`ing traffic control functions for Several connections, So that
`the bandwidth adaptation functions are limited by inad
`equate processing capabilities.
`0022 Many protocols and network management systems
`also perform “windowing of network resources, that is, the
`allocation of transmission periods and bandwidths to SyS
`tems for communication through the network and the adjust
`ment of the windows as traffic needs change. The perfor
`mance of a TCP/IP monitored transmission, for example,
`depends heavily on the value is determined for the threshold
`for window growth, that is, the speed with which a window
`may be increased, and the maximum window size. If a
`window size is too small or the threshold is too low, the
`TCP/IP will not be able to transfer Sufficient data for
`optimum performance and, if the prescribed window Size is
`too large, TCP/IP may lose packets or network congestion
`may ensue, or both.
`0023 For example, under some conditions a current
`window Size may grow to a maximum window Size that may
`exceed the window Size optimal for the link and Such an
`oversize window may allow TCP/IP to transmit excessive
`numbers of packets. The Superfluous packets may exceed the
`optimal number of packets for the available bandwidth and
`buffer Space available at nodes between the Sending and
`receiving end nodes of the connection, thereby overloading
`the System.
`
`0024 Yet other problems occurs in the handling of packet
`dropping under the TCP/IP protocol when excessive packets
`are transmitted. For example, if the slowest intermediate
`node between a Sending end node and a receiving node of a
`network connection has insufficient buffer Space to hold the
`Surplus packets, the packets will be dropped. The dropping
`of packets will result in TCP/IP initiating a time-out and the
`retransmission of packets. Depending on the version of
`TCP/IP, the protocol will either reduce the win