EP1552408A2 - Systeme et procede destines au delestage du tcp/ip independamment d'un produit de retard de largeur de bande - Google Patents

Systeme et procede destines au delestage du tcp/ip independamment d'un produit de retard de largeur de bande

Info

Publication number
EP1552408A2
EP1552408A2 EP03791992A EP03791992A EP1552408A2 EP 1552408 A2 EP1552408 A2 EP 1552408A2 EP 03791992 A EP03791992 A EP 03791992A EP 03791992 A EP03791992 A EP 03791992A EP 1552408 A2 EP1552408 A2 EP 1552408A2
Authority
EP
European Patent Office
Prior art keywords
variables
tcp
host
toe
tcp connection
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
EP03791992A
Other languages
German (de)
English (en)
Other versions
EP1552408A4 (fr
Inventor
Uri Elzur
Frankie Fan
Steve Lindsay
Scott S. Mcdaniel
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Broadcom Corp
Original Assignee
Broadcom Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Broadcom Corp filed Critical Broadcom Corp
Publication of EP1552408A2 publication Critical patent/EP1552408A2/fr
Publication of EP1552408A4 publication Critical patent/EP1552408A4/fr
Withdrawn legal-status Critical Current

Links

Classifications

    • 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
    • H04L12/00—Data switching networks
    • H04L12/28—Data switching networks characterised by path configuration, e.g. LAN [Local Area Networks] or WAN [Wide Area Networks]
    • H04L12/2854—Wide area networks, e.g. public data networks
    • H04L12/2856—Access arrangements, e.g. Internet access
    • H04L12/2869—Operational details of access network equipments
    • H04L12/2898—Subscriber equipments
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00—Data switching networks
    • H04L12/28—Data switching networks characterised by path configuration, e.g. LAN [Local Area Networks] or WAN [Wide Area Networks]
    • H04L12/46—Interconnection of networks
    • H04L12/4633—Interconnection of networks using encapsulation techniques, e.g. tunneling
    • 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
    • H04L47/19—Flow control; Congestion control at layers above the network layer
    • H04L47/193—Flow control; Congestion control at layers above the network layer at the transport layer, e.g. TCP related
    • 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
    • H04L47/24—Traffic characterised by specific attributes, e.g. priority or QoS
    • H04L47/2441—Traffic characterised by specific attributes, e.g. priority or QoS relying on flow classification, e.g. using integrated services [IntServ]
    • 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
    • H04L47/27—Evaluation or update of window size, e.g. using information derived from acknowledged [ACK] packets
    • 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
    • H04L47/28—Flow control; Congestion control in relation to timing considerations
    • H04L47/283—Flow control; Congestion control in relation to timing considerations in response to processing delays, e.g. caused by jitter or round trip time [RTT]
    • 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
    • H04L47/34—Flow control; Congestion control ensuring sequence integrity, e.g. using sequence numbers
    • 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
    • H04L47/37—Slow start
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L63/00—Network architectures or network communication protocols for network security
    • H04L63/14—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic
    • H04L63/1441—Countermeasures against malicious traffic
    • H04L63/1458—Denial of Service
    • 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
    • H04L69/10—Streamlined, light-weight or high-speed protocols, e.g. express transfer protocol [XTP] or byte stream
    • 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
    • H04L69/12—Protocol engines
    • 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
    • H04L69/162—Implementation details of TCP/IP or UDP/IP stack architecture; Specification of modified or new header fields involving adaptations of sockets based mechanisms
    • 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
    • H04L2463/00—Additional details relating to network architectures or network communication protocols for network security covered by H04L63/00
    • H04L2463/141—Denial of service attacks against endpoints in a network

Definitions

  • Certain embodiments of the present invention relate to processing of TCP data and related TCP information. More specifically, certain embodiments relate to a method and system for TCP/IP offload independent of bandwidth delay product.
  • TCP transmission control protocol
  • a transmission control protocol offload engine may be utilized to redistribute TCP processing from the host system onto specialized processors which may have suitable software for handling TCP processing.
  • the TCP offload engines may be configured to implement various TCP algorithms for handling faster network connections, thereby allowing host system processing resources to be allocated or reallocated to application processing.
  • a TCP connection can be offloaded from a host to a dedicated TCP/IP offload engine (TOE). Some of these host resources may include CPU cycles and subsystem memory bandwidth.
  • TCP connection state information is offloaded from the host, for example from a host software stack, to the TOE.
  • a TCP connection can be in any one of a plurality of states at a given time.
  • TCP software may be adapted to manage various TCP defined states. Being able to manage the various TCP defined states may require a high level of architectural complexity in the TOE.
  • Offloading state information utilized for processing a TCP connection to the TOE may not necessarily be the best solution because many of the states such as CLOSING, LAST_ACK and FIN_WAIT_2 may not be performance sensitive. Furthermore, many of these non-performance sensitive states may consume substantial processing resources to handle, for example, error conditions and potentially malicious attacks. These are but some of the factors that substantially increase the cost of building and designing the TOE.
  • a TOE that has control, transferred from the host, of all the state variables of a TCP connection may be quite complex, can use considerable processing power and may require and consume a lot of TOE onboard- memory.
  • the TCP connection offloaded to the TOE that has control, transferred from the host, of all the state variables of the TCP connection can be inflexible and susceptible to connection loss.
  • TCP segmentation is a technology that may permit a very small portion of TCP processing to be offloaded to a network interface card (NIC).
  • NIC network interface card
  • a NIC that supports TCP segmentation does not truly incorporate a full transmission control processing offload engine. Rather, a NIC that supports TCP segmentation only has the capability to segment outbound TCP blocks into packets having a size equivalent to that which the physical medium supports. Each of the outbound TCP blocks are smaller than a permissible
  • TCP window size For example, an Ethernet network interface card that supports TCP Segmentation, may segment a 4KB block of TCP data into 3 Ethernet packets. The maximum size of an Ethernet packet is 1518 bytes inclusive of header and a trailing CRC.
  • a device that supports TCP segmentation does track certain TCP state information such as the TCP sequence number that is related to the data that the offload NIC is segmenting. However, the device that supports TCP segmentation does not track any state information that is related to inbound traffic, or any state information that is required to support TCP acknowledgements or flow control.
  • a NIC that supports full TCP offload in the established state is responsible for handling TCP flow control, and responsible for handling incoming TCP acknowledgements, and generating outbound TCP acknowledgements for incoming data.
  • TCP segmentation may be viewed as a subset of TCP offload.
  • TCP segmentation allows the protocol stack or operating system to pass information in the form of blocks of TCP data that has not been segmented into individual TCP packets to a device driver.
  • the block of data may be greater than the size of an Ethernet packet.
  • the block of data to be segmented could 4 Kbytes or 16 Kbytes.
  • a network adapter associated with the device driver may acquire the blocks of TCP data, packetize the acquired blocks of TCP data into 1518-byte Ethernet packets and update certain fields in each incrementally created packet. For example, the network adapter may update a corresponding TCP sequence number for each of the TCP packets by incrementing the TCP sequence number for each of the packets.
  • IP ID field and flag field would also have to be updated for each packet.
  • IP ID IP identification
  • TCP segmentation may only be done on a block of data that is less than a TCP window size. This is due to the fact that a device implementing TCP segmentation has no influence over TCP flow control. Accordingly, the device implementing TCP flow control only segment outbound TCP packets.
  • a TCP segmentation device does not examine incoming packets and as such, has no influence over flow control. Any received acknowledgement packet is passed up to the host for processing. In this regard, acknowledgement packets that are utilized for flow control are not processed by the TCP segmentation device. Moreover, a TCP segmentation device does not perform congestion control or "slow-start" and does not calculate or modify any variables that are passed back to the operating system and/or host system processor.
  • TCP segmentation Another limitation with TCP segmentation is that information tracked by TCP segmentation is only information that is pertinent for the lifetime of the
  • the TCP segmentation device may track TCP segmentation numbers but not TCP acknowledgement (ACK) numbers. Accordingly, the TCP segmentation device tracks only a minimal subset of information related to corresponding TCP data. This limits the capability and/or functionality of the TCP segmentation device.
  • ACK TCP acknowledgement
  • a further limitation with TCP segmentation is that a TCP segmentation device does not pass TCP processed information back to an operating system and/or host processor. This lack of feedback limits the TCP processing that otherwise may be achieved by an operating system and/or host system processor.
  • a system for TCP/IP offload may include, for example, a host and a TCP/IP offload engine (TOE).
  • the host may be coupled to the TOE.
  • the host may transfer control of at least a portion of TCP connection variables associated with the TCP connection to the TOE.
  • the TOE may update at least a portion of the TCP connection variables and transfer or feedback the updated TCP connection variables back to the host.
  • a system for TCP connection offload.
  • the system may include, for example, a host and a network interface card (NIC) that may be coupled to the host.
  • NIC network interface card
  • control of state information is split between the host and the NIC. Accordingly, information may be transferred to the NIC and the NIC may update at least a portion of the transferred information. Subsequently, the NIC may transfer at least a portion of the updated information back to the host where the host may utilize this information to manage this and/or another connection.
  • the invention may provide a method for TCP/IP offload.
  • the method may include, for example, one or more of the following: deciding to offload a particular TCP connection from a host to a
  • connection variables associated with the particular TCP connection may be transferred back to the host for processing by the host.
  • TCP/IP offload method may include, for example, one or more of the following: deciding to offload an established TCP connection from a host to a TOE; transferring control of segment-variant variables to the TOE from the host; sending a snapshot of segment-invariant variables and connection-invariant variables to the TOE; and independently processing incoming TCP packets via the TOE based upon the segment- variant variables and the snapshot.
  • the TOE may update at least a portion of the segment-variant variables and snapshot and transfer at least portions of the segment-variant variables and the snapshot back to the host.
  • the host may handle all TCP states except possibly for the ESTABLISHED state which may be offloaded to the TOE.
  • the invention may also include a method that processes a TCP connection, which may include, for example, one or more of the following: establishing the TCP connection; sharing a control plane for the TCP connection between a host and a TOE; and communicating updated TCP connection variables from the TOE back to the host. Accordingly, at least a portion of the updated TCP connection variables may be utilized to control the
  • a method for TCP offload may include acquiring TCP connection variables from a host and managing at least one TCP connection using the acquired TCP connection variables. At least a portion of the acquired TCP connection variables may be updated and at least some of the updated TCP connection variables may be transferred back to the host.
  • the TCP connection variables may be independent of bandwidth delay product. At least a portion of the updated TCP connection variables may be utilized by the host to process the TCP connection or another TCP connection.
  • a stack may be utilized to transfer the TCP connection variables between at least the host and a TOE. In this regard, the TOE may pull the
  • the invention may also provide a machine-readable storage, having stored thereon, a computer program having at least one code section for providing TCP offload.
  • the at least one code section may be executable by a machine for causing the machine to perform steps which may include acquiring TCP connection variables from a host and managing at least one TCP connection using the acquired TCP connection variables. At least a portion of the acquired TCP connection variables may be updated and transferred back to the host.
  • the TCP connection variables may be independent of bandwidth delay product.
  • the machine-readable storage may further include code for utilizing at least a portion of the updated TCP connection variables to process the TCP connection or another TCP connection.
  • the machine-readable storage may include code for pulling the TCP connection variables from a stack, code for pushing updated TCP connection variables onto the stack, and code for pulling connection variables from the stack.
  • FIG. 1 is a block diagram of a system that provides TCP/IP offload in accordance with an embodiment of the invention.
  • FIG. 2 is a flow chart illustrating exemplary steps for TCP/IP offloading in accordance with an embodiment of the invention.
  • FIG. 3 is a flow chart illustrating exemplary steps for providing TCP/IP offload in accordance with an embodiment of the invention.
  • FIG. 4 is a flow chart illustrating exemplary steps that may be utilized for TCP offload in accordance with an embodiment of the invention.
  • Certain aspects of the invention may provide a method for TCP offload, which may include acquiring TCP connection variables from a host and managing at least one TCP connection using the acquired TCP connection variables. At least a portion of the acquired TCP connection variables may be updated and at least some of the updated TCP connection variables may be transferred back to the host.
  • the TCP connection variables may be variables that are independent of bandwidth delay product. At least a portion of the updated TCP connection variables may be utilized by the host to process the TCP connection or another TCP connection.
  • a stack may be utilized to transfer the TCP connection variables between at least the host and a TOE. In this regard, the host may push the TCP connection variables onto the stack and the TOE may pull the TCP connection variables from the stack.
  • the updated TCP connection variables may be placed on the stack by the TOE and the host may subsequently pull the updated TCP connection variables from the stack.
  • TCP segmentation each of the outbound TCP blocks are smaller than a permissible TCP window size utilized for TCP segmentation.
  • the invention is not limited in this regard. Accordingly, in an aspect of the invention, a TOE device may have the capability to provide much further TCP processing and offload than a device that simply supports TCP segmentation.
  • Various aspects of the invention may overcome the TCP segmentation limitation in which TCP segmentation may only be done on a block of data that is less than a TCP window size. In this regard, in order to overcome this limitation, in accordance with an aspect of the invention, since the TOE supports management of TCP flow control, the
  • TOE may be adapted to segment large blocks of data down to the individual packets.
  • the TOE may ensure that transmissions where scheduled such that the sender never sent data beyond the TCP window. Additionally, packetization in accordance with an embodiment of the invention may be done beyond the TCP window size.
  • the TOE takes incoming received packets that are acknowledgement packets for the outbound TCP data stream and acknowledges those outbound packets. If the acknowledgement packet causes the window size to increase, then more packets may be sent out by the TOE device in accordance with an aspect of the invention.
  • TCP segmentation is a transmit-only related technology that does limited TCP processing of transmitted packets, the TOE in accordance with various embodiments of the invention is not so limited.
  • the TOE in accordance with an embodiment of the invention may process and manage both transmitted and received packets. Furthermore, a much broader range of TCP processing and management may be done by the TOE in accordance with the invention than with a TCP segmentation device.
  • TCP information may be passed to a NIC from an operating system and/or host system processor in such a manner that the NIC maybe viewed as the owner of the TCP connection.
  • the NIC may then manage and update the TCP state information, which may include but is not limited to, TCP segment numbers and acknowledgment numbers.
  • the processed and/or updated information may be passed back to an operating system and/or host system processor.
  • the host or system processor may then utilize the information passed back to it from the NIC.
  • TCP segmentation does not provide this feedback of information to the host system processor and/or operating system.
  • Certain embodiments of the invention may also provide a robust and efficient transmission control protocol/internet protocol (TCP/IP) offload scheme that may be adapted, for example, to allow the partition of TCP processing between a TCP/IP offload engine (TOE) and a host TCP/IP implementation.
  • the host TCP/IP implementation may include one or more host TCP/IP applications and one or more host processors.
  • the TCP offload scheme may offload the connections that are in an ESTABLISHED state to the TOE.
  • aspects of the invention may include the offloading of corresponding TCP state variables that may be utilized, for example, during the ESTSABLISHED state.
  • the TCP/IP offload scheme may split a TCP control plane between the host software and the TOE.
  • the TOE may be designed, for example, to implement a subset or a minimum subset of the TCP control plane which may be less complex to implement and may utilize less memory.
  • the TOE which may be adapted to such an offload scheme, may be implemented in a cost effective manner. The more complicated aspects of TCP connection management may be handled, for example, by the host software and may provide greater reliability and flexibility.
  • FIG. 1 is a block diagram of a system that provides TCP/IP offload in accordance with an embodiment of the invention.
  • the system may include, for example, a host 10, host application software 12 and a TOE 20.
  • the host 10 may include, for example, a host CPU 30 and a host memory 40.
  • the host memory 40 may be adapted to include, for example, an application buffer 50.
  • the application buffer 50 may be adapted to include, for example, a transmission application buffer (TxBuf) 60 and a receive application buffer (RxBuf) 70.
  • the TOE 20 may include, for example, a direct memory access (DMA) engine 25 and a FIFO buffer 70.
  • DMA direct memory access
  • the host 10 may be coupled to the TOE 20 via a host interface 80.
  • the host interface may include, but is not limited to a peripheral component interconnect (PCI) bus, PCI-X bus, ISA, SCSI or any other suitable bus.
  • PCI peripheral component interconnect
  • the TOE 20 may be coupled to a physical communications medium 90.
  • the physical communication medium 90 may be a wired medium, wireless medium or a combination thereof.
  • the physical communication medium 90 may include, but is not limited to, Ethernet and fibre channel.
  • the host 10 may be, at least in part, disposed on a network interface card (NIC) that includes the TOE 20. Accordingly, in an aspect of the invention, the TCP state plane may be split between the host 10 and the TOE 20.
  • NIC network interface card
  • a TCP connection may be completely described, for example, by three different sets of variables.
  • the three sets of variables may be, for example, connection-invariant variables, segment-invariant variables and segment-variant variables.
  • the connection-invariant variables may be constant during the lifetime of the TCP connection.
  • the segment- invariant variables may not change from TCP segment to TCP segment, but may change from time to time during the lifetime of the TCP connection.
  • the segment-variant variables may change from TCP segment to TCP segment.
  • Connection-invariant variables may include, for example, source IP address, destination IP address, IP time-to-live (TTL), IP type-of-service
  • TOS source TCP port number
  • destination TCP port number initial send sequence number
  • initial receive sequence number initial receive sequence number
  • send window scaling factor receive window scaling factor
  • Segment-invariant variables may include, but are not limited to, source MAC address, next hop's MAC address, MAC layer encapsulation, effective maximum segment size, keep-alive intervals and maximum allowance and flags such as, for example, nagle algorithm enable and keep-alive enable.
  • Segment-variant variables may include, but are not limited to, IP packet identifier; send and receive sequence variables such as, for example, sequence number for first un-acked data (SND_UNA), sequence number for next send (SND_NXT), maximum sequence number ever sent (SND_MAX), maximum send window (MAX_WIN), sequence number for next receive
  • SND_UNA sequence number for first un-acked data
  • SND_NXT sequence number for next send
  • SND_MAX maximum sequence number ever sent
  • MAX_WIN maximum send window
  • Additional exemplary segment-variant variables may include congestion window variables such as congestion window (SND_CWIN) and slow start threshold (SSTHRESH) round trip time variables which may include, but are not limited to, smoothed round trip time (RTT) and smoothed delta (DELTA).
  • congestion window variables such as congestion window (SND_CWIN) and slow start threshold (SSTHRESH) round trip time variables which may include, but are not limited to, smoothed round trip time (RTT) and smoothed delta (DELTA).
  • RTT smoothed round trip time
  • DELTA smoothed delta
  • Other exemplary segment-variant variables may include time remaining for retransmission, time remaining for delay acknowledgement, time remaining for keep alive, time remaining for PUSH and TCP state and timestamp.
  • FIG. 2 is a flow chart illustrating exemplary steps for TCP/IP offloading in accordance with an embodiment of the invention.
  • the host software may transfer control of the segment-variant variables to the TOE 20. In one example, a portion of the host software protocol control block or TCP control block may be transferred to the TOE 20.
  • the host software may take a snapshot of the remaining variables such as the connection-invariant variables and/or the segment invariant variables and send the snapshot to the TOE 20. In one example, the snapshot may be used over and over again by the TOE 20.
  • the host software may post a buffer in the host memory 40. For example, the host software may post the application buffer 50 in the host memory 40 and may set up the transmit application buffer
  • the TOE 20 may be responsible for managing the complete TCP connection, including, for example, segmentation, acknowledgement processing, windowing and congestion avoidance.
  • at least a portion of the variables that have been updated may be transferred back to the host for processing.
  • the TOE 20 may process or independently process, incoming TCP segments from the physical communications medium 90 and may place at least a portion such as a payload, of the incoming TCP segments into the host memory 40 via the DMA engine 25.
  • the incoming TCP segment payload may be placed in the RX application buffer 70 portion of the application buffer 50 via the DMA engine 25.
  • the TOE 20 may have exclusive read-write access to offloaded segment-variant variables and may exclusively update the offloaded segment-variant variables.
  • the host software or host application software 12 may have read-write access to the segment-invariant variables.
  • the TOE 20 may have read-only access to the segment-invariant variables. If the host application software 12 changes the variables such as the next hop's MAC address, the host application software 12 may notify the TOE 20 by, for example, sending a message to the TOE 20. The TOE 20 may then update the variables. The updated variables may be fed back to the host application software 12 where they may be utilized for TCP processing, for example. Accordingly, the connection-invariant variables may exist in both the host software and the TOE 20.
  • FIG. 3 is a flow chart illustrating exemplary steps for providing TCP/IP offload in accordance with an embodiment of the invention.
  • the host 10 may determine whether one or more of the connection variables such as the segment-invariant variables controlled by the host 10 have changed. For example, the host software may change one or more variables such as a next hop MAC address. If one or more of the connection variables controlled by the host 10 are not changed, then the process may be complete. If one or more of the connection variables controlled by the host 10 are changed, then, in step 304, the host software may notify the TOE 20 of the change in the one or more connection variables controlled by the host 10. In step 306, the TOE 20 may accordingly update one or more of the variables. In step 308, the TOE may pass the updated variables back to the host 10.
  • the connection variables such as the segment-invariant variables controlled by the host 10 have changed. For example, the host software may change one or more variables such as a next hop MAC address. If one or more of the connection variables controlled by the host 10 are not changed, then the
  • Some embodiments according to the present invention may include one or more of the following advantages. Some embodiments may be more reliable and may provide for the uploading of connection from the TOE to the host and offloading of connections from the host to the TOE at any time. Since less state information may be kept by the TOE hardware, uploading and offloading, for example, selected connections can be accelerated. An offloaded connection may be uploaded by returning control of, for example, the segment-variant variables corresponding to the offloaded connection back to the host 10. The uploaded connection may subsequently be offloaded by transferring, for example, the control of the segment-variant variables corresponding to the uploaded connection to the TOE 20.
  • FIG. 4 is a flow chart illustrating exemplary steps that may be utilized for TCP offload in accordance with an embodiment of the invention.
  • a TOE may acquire or receive variables that are independent of the bandwidth delay product from a host system.
  • the TOE may manage the connection utilizing the acquired or received variables that are independent of the bandwidth delay product.
  • the TOE may update at least a portion of the acquired variables that are independent of the bandwidth delay product.
  • at least a portion of the updated variables that are independent of the bandwidth may be transferred back to the host.
  • the host may utilize the updated variables that are independent of the bandwidth delay product that have been transferred to it for TCP processing.
  • a stack 14 may be utilized to facilitate the transfer of the variables that are independent of the bandwidth delay product.
  • the stack 14 may be implemented in hardware, software or a combination thereof.
  • the TOE may be adapted to pull information from the stack 14 and to push updated information onto the stack 14.
  • the host may also be adapted to push TCP information onto the stack 14 and to pull the updated information from the stack 14.
  • the TOE may pull the variables that are independent of the bandwidth delay product from the stack 14.
  • the updated variables that are independent of the bandwidth delay product may be pushed onto the stack 14.
  • the host may then pull the updated variables that are independent of the bandwidth delay product from the stack 14.
  • the TOE may provide a more flexible approach to TCP processing compared to a TCP Segmentation offload deice, since the TOE device may facilitate TCP processing on both the received side and the transmit side.
  • the TOE may be adapted to handle receive and transmit variables, the TOE provides a more flexible and efficient methodology for supporting the efficient setup and tear down of network connections.
  • Certain embodiments of the invention may offer better resistance against denial-of-service (DoS) attacks or other attacks as connection setup may be handled by a host that is more flexible and more powerful than the TOE NIC.
  • DoS denial-of-service
  • an attacker attempts to consume as many resources on the targeted or attacked system, thereby preventing the targeted system from providing services to other network devices.
  • the frequent introduction of new attacks may make a flexible host with sufficient memory and CPU power a better choice for running connection setup.
  • the flexible host may be a better choice than, for example, a particular hardware TOE that may have limited code space, computer power, system knowledge and flexibility.
  • the decision to honor a connection request may, at times, be based upon, for example, sophisticated and dynamic heuristics.
  • the TOE NIC may be more efficient in handling, for example, connections that are in performance sensitive states of the TCP state machine.
  • the TOE NIC may be adapted to upload connections that are no longer in performance sensitive states and to offload connections that are in performance sensitive states. Such actions may positively impact such figures of merit such as, for example, hardware TOE efficiency.
  • Other aspects of the invention may be more efficient and may provide better over all system performance because, for example, the host may use flexible, changing, easy-to-update, easy-to- upgrade and more sophisticated algorithms to decide which connections to offload or to upload.
  • Some embodiments according to the present invention may provide statistics to the host relating to resource utilization.
  • the statistics may include, for example, one or more of the following: available resources; utilization of bandwidth per offloaded connection; number of frames per offloaded connection; errors per offloaded connection; change of state of a transport layer protocol (TLP) such as, for example, TCP, or an upper layer protocol (ULP); trend of utilization such as uptake in rate, slow down, for example; and resource consumption per offloaded connection.
  • TLP transport layer protocol
  • ULP upper layer protocol
  • trend of utilization such as uptake in rate, slow down, for example
  • resource consumption per offloaded connection The host may use the statistical information at its own discretion to help drive the upload or offload decision process. For example, the host may utilize the statistical information to upload some connections while offloading others.
  • the host may also contemplate other criteria such as modes of operation, computation or network load profiles, presently executed applications and roles in the network, for example. Some of these criteria may be dynamic criteria.
  • Certain embodiments of the invention may also provide fail-over support from a failed TOE NIC to an operating TOE NIC. Fail-over may include, for example, designating a NIC as having failed when the network cable is unplugged from the network or any other failure of an existing network link. Thus, even though the hardware of one TOE NIC may fail, the connection may still be maintained by transferring state information associated with the failed TOE NIC to another functional TOE NIC. The robustness of the transfer may be further enhanced by part of the connection state information being maintained by the host and part of the connection state information being maintained by the TOE NIC.
  • the present invention may be realized in hardware, software, or a combination of hardware and software.
  • the present invention may be realized in a centralized fashion in one computer system or in a distributed fashion where different elements are spread across several interconnected computer systems. Any kind of computer system or other apparatus adapted for carrying out the methods described herein is suited.
  • a typical combination of hardware and software may be a general-purpose computer system with a computer program that, when being loaded and executed, controls the computer system such that it carries out the methods described herein.
  • the present invention also may be embedded in a computer program product, which comprises all the features enabling the implementation of the methods described herein, and which when loaded in a computer system is able to carry out these methods.
  • Computer program in the present context means any expression, in any language, code or notation, of a set of instructions intended to cause a system having an information processing capability to perform a particular function either directly or after either or both of the following: a) conversion to another language, code or notation; b) reproduction in a different material form.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Computer Security & Cryptography (AREA)
  • Computer Hardware Design (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

Des modes de réalisation de l'invention concernent un délestage du TCP (protocole de commande de transmission) pouvant consister en une acquisition de variables de connexion TCP à partir d'un hôte et en une gestion d'au moins une connexion TCP au moyen des variables de connexion TCP acquises. Au moins une partie des variables de connexion TCP acquises peut être mise à jour et au moins une partie des variables de connexion TCP mises à jour peut être retransférée vers l'hôte. Dans un mode de réalisation de l'invention, les variables de connexion TCP peuvent être des variables indépendantes du produit de retard de la largeur de bande. Au moins une partie des variables de connexion TCP mises à jour peut être utilisée par l'hôte aux fins de traitement de la connexion TCP ou d'une autre connexion TCP. L'hôte peut pousser les variables sur l'empilement et le TOE (moteur de délestage TCP/IP) peut retirer les variables de l'empilement. Des variables de connexion TCP mises à jour peuvent également être poussées sur l'empilement par le TOE et être retirées de celui-ci par l'hôte.
EP03791992A 2002-08-30 2003-08-29 Systeme et procede destines au delestage du tcp/ip independamment d'un produit de retard de largeur de bande Withdrawn EP1552408A4 (fr)

Applications Claiming Priority (5)

Application Number Priority Date Filing Date Title
US40716502P 2002-08-30 2002-08-30
US407165P 2002-08-30
US40861702P 2002-09-06 2002-09-06
US408617P 2002-09-06
PCT/US2003/027351 WO2004021150A2 (fr) 2002-08-30 2003-08-29 Systeme et procede destines au delestage du tcp/ip independamment d'un produit de retard de largeur de bande

Publications (2)

Publication Number Publication Date
EP1552408A2 true EP1552408A2 (fr) 2005-07-13
EP1552408A4 EP1552408A4 (fr) 2010-10-06

Family

ID=31981477

Family Applications (1)

Application Number Title Priority Date Filing Date
EP03791992A Withdrawn EP1552408A4 (fr) 2002-08-30 2003-08-29 Systeme et procede destines au delestage du tcp/ip independamment d'un produit de retard de largeur de bande

Country Status (3)

Country Link
EP (1) EP1552408A4 (fr)
CN (1) CN100363922C (fr)
WO (1) WO2004021150A2 (fr)

Families Citing this family (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8549170B2 (en) * 2003-12-19 2013-10-01 Nvidia Corporation Retransmission system and method for a transport offload engine
US7533176B2 (en) 2004-07-14 2009-05-12 International Business Machines Corporation Method for supporting connection establishment in an offload of network protocol processing
US7493427B2 (en) 2004-07-14 2009-02-17 International Business Machines Corporation Apparatus and method for supporting received data processing in an offload of network protocol processing
US7930422B2 (en) 2004-07-14 2011-04-19 International Business Machines Corporation Apparatus and method for supporting memory management in an offload of network protocol processing
US20070033301A1 (en) * 2005-07-18 2007-02-08 Eliezer Aloni Method and system for transparent TCP offload with dynamic zero copy sending
TWI290799B (en) 2005-12-16 2007-12-01 Ind Tech Res Inst Extensible protocol processing system
DE602007013652D1 (de) 2006-08-04 2011-05-19 Canon Kk Kommunikationsvorrichtung und Kommunikationssteuerungsverfahren
EP2079203A1 (fr) * 2008-01-08 2009-07-15 Axis AB Déchargement de réseau avec perte de paquet réduite
CN104601484B (zh) * 2015-01-20 2017-10-31 电子科技大学 一种tcp卸载引擎的发送单元
WO2019005092A1 (fr) * 2017-06-30 2019-01-03 Intel IP Corporation Dispositifs de déchargement partiel des tâches de traitements de protocole
CN109714302B (zh) * 2017-10-25 2022-06-14 阿里巴巴集团控股有限公司 算法的卸载方法、装置和系统
CN110830381B (zh) 2018-08-10 2021-10-26 华为技术有限公司 拥塞控制方法及相关设备
CN110109852B (zh) * 2019-04-03 2020-11-24 华东计算技术研究所（中国电子科技集团公司第三十二研究所） 硬件实现tcp_ip协议的方法

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6427171B1 (en) * 1997-10-14 2002-07-30 Alacritech, Inc. Protocol processing stack for use with intelligent network interface device
US6434620B1 (en) * 1998-08-27 2002-08-13 Alacritech, Inc. TCP/IP offload network interface device

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5937169A (en) * 1997-10-29 1999-08-10 3Com Corporation Offload of TCP segmentation to a smart adapter
KR100645424B1 (ko) * 2000-08-07 2006-11-14 삼성전자주식회사 모뎀 및 그 제어방법
US7496689B2 (en) * 2002-04-22 2009-02-24 Alacritech, Inc. TCP/IP offload device
US6968358B2 (en) * 2002-07-25 2005-11-22 International Business Machines Corporation Method and apparatus for network communication card memory management

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6427171B1 (en) * 1997-10-14 2002-07-30 Alacritech, Inc. Protocol processing stack for use with intelligent network interface device
US6434620B1 (en) * 1998-08-27 2002-08-13 Alacritech, Inc. TCP/IP offload network interface device

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Eric Yeh, Herman Chao, Venu Mannem, Joe Gervais, Bradley Booth: "Introduction to TCP/IP Offload Engine (TOE)"[Online] 1 April 2002 (2002-04-01), XP002596417 10GEA - 10 Gigabit Ethernet Alliance Retrieved from the Internet: URL:http://ethernetalliance.emergeinteractive.com/files/static_page_files/D14863B2-1D09-3519-AD64EB84C8B58D4A/TCIP%20Offload%20Engine.pdf> [retrieved on 2010-08-13] *
See also references of WO2004021150A2 *

Also Published As

Publication number Publication date
CN1679015A (zh) 2005-10-05
WO2004021150A3 (fr) 2004-08-12
CN100363922C (zh) 2008-01-23
WO2004021150A2 (fr) 2004-03-11
EP1552408A4 (fr) 2010-10-06

Similar Documents

Publication Publication Date Title
US7313623B2 (en) System and method for TCP/IP offload independent of bandwidth delay product
EP1513321B1 (fr) Système et procédé destinés au délestage du tcp/ip indépendamment d'un produit de retard de largeur de bande
US8064459B2 (en) Method and system for transparent TCP offload with transmit and receive coupling
US8174975B2 (en) Network adapter with TCP support
US7912064B2 (en) System and method for handling out-of-order frames
TWI411279B (zh) 封包聚合的方法與系統
US7613109B2 (en) Processing data for a TCP connection using an offload unit
US7420931B2 (en) Using TCP/IP offload to accelerate packet filtering
US7512144B2 (en) Method and system for transmission control protocol (TCP) retransmit processing
US20060268710A1 (en) Detecting change in a transport protocol window size without data transmission
US20070297334A1 (en) Method and system for network protocol offloading
US20070223529A1 (en) Methods and apparatus for estimating bandwidth of a data network
JP2003333076A (ja) ネットワークスタックをオフロードする方法
EP1552408A2 (fr) Systeme et procede destines au delestage du tcp/ip independamment d'un produit de retard de largeur de bande
KR101067394B1 (ko) 페일오버 이벤트를 지원하는 네트워크 상태 객체의 다중오프로드용 방법 및 컴퓨터 프로그램 제품
US8578040B2 (en) Method, system and article for client application control of network transmission loss tolerance
EP1460804A2 (fr) Système et procédé de gestion des trames des données hors-ordre (FKA réception des trames DES TCP/IP hors-ordre sans copie service)
KR20080042764A (ko) 데이터 네트워크의 대역폭 추정 방법 및 네트워크 노드

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

17P Request for examination filed

Effective date: 20050330

AK Designated contracting states

Kind code of ref document: A2

Designated state(s): AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IT LI LU MC NL PT RO SE SI SK TR

RBV Designated contracting states (corrected)

Designated state(s): DE FR GB

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: BROADCOM CORPORATION

A4 Supplementary search report drawn up and despatched

Effective date: 20100907

17Q First examination report despatched

Effective date: 20120430

17Q First examination report despatched

Effective date: 20120503

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN

18D Application deemed to be withdrawn

Effective date: 20170301