WO2001017153A2 - Scalable server architecture based on asymmetric 3-way tcp - Google Patents
Scalable server architecture based on asymmetric 3-way tcp Download PDFInfo
- Publication number
- WO2001017153A2 WO2001017153A2 PCT/KR2000/000973 KR0000973W WO0117153A2 WO 2001017153 A2 WO2001017153 A2 WO 2001017153A2 KR 0000973 W KR0000973 W KR 0000973W WO 0117153 A2 WO0117153 A2 WO 0117153A2
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- end server
- client
- packet
- pseudo
- server
- Prior art date
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
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/14—Session management
-
- 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
- H04L69/30—Definitions, standards or architectural aspects of layered protocol stacks
- H04L69/32—Architecture of open systems interconnection [OSI] 7-layer type protocol stacks, e.g. the interfaces between the data link level and the physical level
- H04L69/322—Intralayer communication protocols among peer entities or protocol data unit [PDU] definitions
- H04L69/329—Intralayer communication protocols among peer entities or protocol data unit [PDU] definitions in the application layer [OSI layer 7]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
- H04L67/1036—Load balancing of requests to servers for services different from user content provisioning, e.g. load balancing across domain name servers
Definitions
- This invention relates to client-server network processing. More specifically, this invention relates to a server architecture based on asymmetric 3 -way TCP.
- Computers on the Internet communicate with each other by sending packets according to a standard protocol suite including the TCP/IP (Transmission Control Protocol and Internet Protocol) suite.
- TCP/IP Transmission Control Protocol and Internet Protocol
- a scalable server system whose handling capacity can increase in proportion to the frequency of service requests from clients.
- One way to implement such a scalable server system is to employ a high- performance single-processor server. But, this is not an economically desirable solution because one must pay a stiff price for such a single-processor high-performance system.
- Another way is to employ multiple servers to distribute service requests from clients. However, adding servers does not necessarily translate to increased performance because the existing protocol limits handling of service requests to only those servers specified by the clients, thereby creating a bottleneck.
- an asymmetric 3- way TCP (ATCP) scheme that includes a front-end server and a back-end server, both communicating with a client.
- the front-end server receives a service request from the client and sends a service command necessary to process the service request to the back- end server.
- the back-end server executes the service command and sends the result back directly to the client in a pseudo TCP packet which would be recognized by the client as being sent by the front-end server.
- the pseudo TCP packet has the IP address and the port number of the front-end server to look like a packet sent by the front-end server.
- FIG. 1 is an illustration of an asymmetric 3-way TCP (ATCP) scheme of the present invention.
- FIG. 2 is a schematic diagram of the ATCP of the present invention.
- FIG. 3 is a flowchart of a software routine run at the front-end server.
- FIG. 4 is a flowchart of a software routine run at the back-end server.
- FIG. 5 is a flowchart of a software routine run by the front-end server to send a service command the back-end server through an ATCP_command_back-end_server packet.
- FIG. 6 is a flowchart of a software routine run at the back-end server upon receiving the ATCP_command_to_back-end_server packet from the front-end server.
- FIG. 7 is a flowchart of a software routine run at the back-end server before sending an ATCP_service_reply packet to the client and an ATCP_ACK packet to the front-end server.
- FIG. 8 is a flowchart of a software routine run at the front-end server upon receiving the ATCP_ACK packet from the back-end server.
- FIG. 9 is a flowchart of a software routine run at the front-end server upon receiving an ACK packet from the client.
- FIG. 10 is a flowchart of a software routine run at the back-end server upon receiving the ATCP_ACK packet from the front-end server.
- FIG. 11 is a block diagram for a back-end server that assembles a pseudo TCP packet.
- FIG. 12 is a state transition diagram of a state machine controlling the back-end server.
- FIG. 1 illustrates the operations of a client-server system based on Asymmetric TCP (ATCP) of the present invention.
- the system has a client 11 communicating with a front-end server 12 and a back-end server 13.
- the operations may be summarized as follows:
- the client makes a service request to the front-end server through the socket of the front-end server whose TCP connection is specified by the client. Note that the client does not specify any connection with the back-end server and does not even know the presence of the back-end server.
- the front-end server relays the service request received from the client to the back-end servers together with the source/destination port numbers and source/destination IP addresses and the acknowledge number expected by the client and the sequence number of the TCP packet that would contain the result of the service request.
- the front-end server may simply relay the client's request to the back-end server or sends a specific command after decoding a request. If the front-end server directly handles the client's service request, the front-end server would send the result to the clients through a conventional TCP packet.
- the back-end server sends a TCP packet containing the service result to the client through the client's socket.
- the TCP packet is made by the back-end server, the packet is disguised as if it were sent by the front-end server by using the destination/source sockets specified by the client and the front-end server. For this reason, the TCP packet sent by the back-end server is referred to as a "pseudo TCP packet.”
- the client sends a TCP packet responding to the service received to the front- end server and sends a further service request to the front-end server, if necessary.
- FIG. 2 shows a schematic diagram of the TCP and ATCP packets exchanged between the client, the front-end server, and the back-end server.
- SYN 21, SYN ACK 22, ACK 23 are conventional TCP packets for handshaking between the client and the front-end server to establish a TCP connection.
- PSH 24 is a conventional TCP packet containing a service request from the client to the front-end server.
- PSH' 24' is a conventional TCP packet that would have been generated by the front-end server to send the front-end server's reply to the service request.
- PSH' 24' is not generated. Instead, the front-end server relays the service request to the back-end server so that the reply to the service requests is handled by the back-end server.
- the front-end server can distribute the processing among multiple processors, thereby enabling a high- performance server system.
- ATCP_command_to_back-end_server 25 of the present invention is not a conventional TCP packet, but an ATCP packet containing a service command from the front-end server to the back-end server, the source/destination IP addresses of the client requesting the service, the port number of the client, the acknowledgement number expected by the client, and the sequence number expected by the client.
- the back-end server receiving receives a service command from the front- end server through ATCP_command_to_back-end_server 25, processes the command and sends the result directly to the client in ATCP_service_reply 26.
- the ATCP_service_reply packet For ATCP_service_reply 26 to be recognized by the client process through the client's TCP layer, the ATCP_service_reply packet must look as if it were sent by the front-end server. This is accomplished by making the destination port number of the ATCP packet 26 assigned the client's port number and the source port number of the ATCP packet assigned the front-end server's port number. The destination port number and the acknowledgement number of the ATCP_service_reply packet 26 are assigned the sequence number and the acknowledgement number received through the ATCP_command_to_back-end_server 25 packet.
- the back-end server After sending ATCP_service_reply 26 to the client, the back-end server sends an acknowledgement to the front-end server through ATCP_ACK 27. Without this acknowledgement, the front-end server has no way of knowing whether the back-end server correctly has processed the command.
- the TCP layer of the front-end server sets the acknowledgement number expected from the client as the sequence number sent to the back-end server plus the size of data sent to the client by the back-end server.
- (g) Receiving ATCP_service_reply 26, the client recognizes it as a packet received by the front-end server and sends an acknowledgement ACK 28, a conventional TCP packet containing an acknowledgement number, to the front-end server.
- the acknowledgement number would be equal or less than the acknowledgement number expected by the front-end server because the front-end server already updated the expected acknowledgement number in step (f). If the expected acknowledgement number had not been updated in step (f), a TCP packet having the response larger than the expected number would have been discarded.
- the front-end server relays the TCP packet received from the client through the established connection between the client and the front-end server. Note, however, that the back-end server need not and does not establish a TCP connection with the client because the back-end server sends the result in a pseudo TCP packet that would look like a TCP packet sent from the front-end server. In fact, the client does not even know of the presence of the back-end servers.
- the packets exchanged by the front-end server and back-end server need not be a TCP packet or of the same format as a TCP packet as long as the packets support a connection-oriented service.
- FIG. 3 is a flowchart of a software routine run at the front-end server.
- the process at the front-end server first initializes the socket connection with the back-end server (step 101), and blocks itself while waiting a service request from a client (step 102).
- the process at the front-end server is awakened, and the process delivers a service command to the back-end server through the ATCP layer.
- the process at the front-end server repeats steps 102-104.
- Processing point (A) represents the point in time when the process run at the front-end server sends a command to the back-end server through the ATCP layer.
- FIG. 4 is a flowchart of a software routine run at the back-end server.
- the process run at the back-end server first initializes the socket connection (step 112) and blocks itself while waiting for a command from the front-end server, (step 113).
- the process is awakened when ATCP_command_to-back-end_server 25, a pseudo TCP packet containing a command from the front-end server, is sent to the ATCP level of the back-end server (step 114).
- ATCP_command_to-back-end_server 25 a pseudo TCP packet containing a command from the front-end server, is sent to the ATCP level of the back-end server (step 114).
- processing point (C) represents the point in time when the command is delivered from the back-end server software to the ATCP level at the back-end server through the API such as a socket.
- FIG. 5 is a flowchart of a software routine run at the front-end server after step
- FIG. 6 is a flowchart of a software routine run at the back-end server.
- the back-end server runs a back-end server command process for executing the command received from the front-end server.
- the software at the back-end server has the ATCP layer that sets up a 3-way ATCP connection (step 109) and wakes up the back-end server command process blocked while waiting for a command from the front-end server (step 110).
- FIG. 7 is a flowchart of a software routine run at the back-end server before sending ATCP_service 26 to the client and ATCP_ACK 27 to the front-end server.
- the result of the service command is received from the back-end server command process, it is sent to the client by incorporating the IP address and port number of the front-end server.
- FIG. 8 is a flowchart of a software routine run at the front-end server upon receiving ATCP_ACK 27 from the back-end server at the time indicated by Point D.
- the front-end server checks whether the ATCP connection is valid (step 120). If not valid, the received packet is discarded (step 121). If valid, the front-end server extracts the size of data transmitted by the back-end server to the client (step 122) and updates the expected acknowledgement by adding the data size (step 123).
- FIG. 9 is a flowchart of a software routine run at the front-end server upon receiving ACK 28 from the client at the time indicated by Point E.
- the front-end server checks whether the ATCP connection is valid (step 124). If not valid, the received packet is discarded (step 126). If valid, it is checked whether the received ACK number is equal or less than the sequence number (step 125). If the received ACK number is larger, the received packet is discarded (step 126). If the received ACK number is equal or less, the ATCP buffer is emptied by the received amount.
- FIG. 10 is a flowchart of a software routine run at the back-end server at receiving ATCP_ACK 29 that was relayed by the front-end server receiving ACK 28 from the client.
- Point F in FIG. 10 corresponds to point F in FIG. 9, the time when the back-end server received ATCP_ACK 29 from the front-end server.
- the back- end server checks whether the ATCP connection is valid (step 129). If not valid, the received packet is discarded (step 131). If valid, it is checked whether the received ACK number is equal or less than the sequence number (step 130). If the received ACK number is larger, the received packet is discarded (step 131). If the received ACK number is equal or less, the ATCP buffer is emptied by the received amount.
- FIG. 11 shows the back-end server which comprises a back-end server controller 30, an NIC(network interface card) 40, a disk 50, a CD 60 and memory 70.
- the NIC functions as an interface between the back-end server and the network cabling.
- the back-end server controller 30 is preferably a program controlling various controlling operations in the back-end server.
- the back-end server controller 30 further comprises a state machine 31 for controlling the back-end server, a NIC controller 32, a command decoder 33, an ALU 34, a disk controller 35, a CD controller 36, a pseudo TCP packetizer 37.
- the NIC controller 32 is a program module for controlling receipt and transmission of packets from and to the network by reading and writing an NIC register.
- the command decoder 33 is a program module to decode the command from the front- end server included in the TCP packet received through the NIC.
- the ALU 34 performs arithmetic and logic operations necessary for performing a command.
- the disk controller 35 is a program module to control the input and output to and from the disk 50.
- the CD controller 36 is a program module to control the output from a CD from the CD 60.
- the pseudo TCP packetizer 37 assembles a pseudo TCP packet by making up the TCP header and the IP header to look like that the packet was sent by the front-end server.
- FIG. 12 is a state transition diagram of a state machine controlling the back-end server of the present invention.
- the back-end server is ready to receive a command from the front-end server through the NIC.
- the received command is decoded by the command decoder 33 to execute the command or to perform I/O.
- the result is incorporated by a pseudo TCP packetizer 37 into a pseudo TCIP/TP packet that looks like a packet sent by the front-end server. The above process is repeated.
- the asymmetric 3-way TCP (ATCP) scheme of the present invention realizes an economic and scalable server system where the server distributes a client request to one or more back-end servers which perform the work and deliver the result to the client.
- ATCP asymmetric 3-way TCP
- the ATCP scheme of the present invention can increase the performance of a web server by replacing the existing TCP layer with the ATCP layer so as to distribute the work among multiple back-end servers.
- the existing web servers may be modified in such a way that the web servers that received the requests from clients may instructs the multiple back-end servers to perform the work and relate the work to the clients.
- back-end servers are transparent to the client in that there is no need for a client to have any prior knowledge of back-end servers and there is no need to change the IP or the port address of the existing servers. Only the front-end server needs to have prior knowledge of the back-end servers.
- the prevent invention was described in terms of a preferred embodiment using the TCP, an alternative embodiment is possible using the UDP (User Datagram Protocol), another transport-layer protocol offered as part of the TCP/IP suite.
- UDP User Datagram Protocol
- the TCP provides a connection-oriented service where a connection is established first before packets are sent in a continuous stream
- the UDP provides a connectionless service where individual datagrams are sent without first establishing a connection.
- the UDP is typically used where guaranteed delivery is not critical, such as video or audio transmission.
- the front-end servers sends to the back-end server the UDP header information (the front-end server's
- the back-end server then assembles a pseudo UDP/IP packet using the information received.
Abstract
Description
Claims
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CA002387400A CA2387400A1 (en) | 1999-09-01 | 2000-08-29 | Scalable server architecture based on asymmetric 3-way tcp |
DE60012007T DE60012007T2 (en) | 1999-09-01 | 2000-08-29 | TRIPLE TCP-BASED SCALABLE SERVER ARCHITECTURE |
AU70388/00A AU7038800A (en) | 1999-09-01 | 2000-08-29 | Scalable server architecture based on asymmetric 3-way tcp |
EP00958987A EP1177500B1 (en) | 1999-09-01 | 2000-08-29 | Scalable server architecture based on asymmetric 3-way tcp |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR9936874 | 1999-09-01 | ||
KR1999/36874 | 1999-09-01 |
Publications (2)
Publication Number | Publication Date |
---|---|
WO2001017153A2 true WO2001017153A2 (en) | 2001-03-08 |
WO2001017153A3 WO2001017153A3 (en) | 2001-12-06 |
Family
ID=19609648
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/KR2000/000973 WO2001017153A2 (en) | 1999-09-01 | 2000-08-29 | Scalable server architecture based on asymmetric 3-way tcp |
Country Status (1)
Country | Link |
---|---|
WO (1) | WO2001017153A2 (en) |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5513314A (en) * | 1995-01-27 | 1996-04-30 | Auspex Systems, Inc. | Fault tolerant NFS server system and mirroring protocol |
-
2000
- 2000-08-29 WO PCT/KR2000/000973 patent/WO2001017153A2/en active IP Right Grant
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5513314A (en) * | 1995-01-27 | 1996-04-30 | Auspex Systems, Inc. | Fault tolerant NFS server system and mirroring protocol |
Also Published As
Publication number | Publication date |
---|---|
WO2001017153A3 (en) | 2001-12-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7483967B2 (en) | Scalable server architecture based on asymmetric 3-way TCP | |
US7685287B2 (en) | Method and system for layering an infinite request/reply data stream on finite, unidirectional, time-limited transports | |
US10284626B2 (en) | Transporting operations of arbitrary size over remote direct memory access | |
US7117267B2 (en) | System and method for providing tunnel connections between entities in a messaging system | |
US20070168545A1 (en) | Methods and devices for processing incomplete data packets | |
US9178966B2 (en) | Using transmission control protocol/internet protocol (TCP/IP) to setup high speed out of band data communication connections | |
US20020099827A1 (en) | Filtering calls in system area networks | |
US8793384B2 (en) | Recovery of disconnected channels over a reliable protocol | |
WO2004062234A1 (en) | Message transfer using multiplexed connections in an osi-tp environment | |
US20060031527A1 (en) | Expanded transmission control protocol, methods of operation and apparatus | |
CA2538602A1 (en) | Internet protocol optimizer | |
US7039710B2 (en) | Method and apparatus for providing multi-client support in a SIP-enabled terminal | |
US6757736B1 (en) | Bandwidth optimizing adaptive file distribution | |
US8539112B2 (en) | TCP/IP offload device | |
US7523179B1 (en) | System and method for conducting direct data placement (DDP) using a TOE (TCP offload engine) capable network interface card | |
EP1177500B1 (en) | Scalable server architecture based on asymmetric 3-way tcp | |
US7213074B2 (en) | Method using receive and transmit protocol aware logic modules for confirming checksum values stored in network packet | |
WO2001017153A2 (en) | Scalable server architecture based on asymmetric 3-way tcp | |
US8639828B1 (en) | Accelerated TCP network communications | |
WO2018142866A1 (en) | Transfer device, transfer method and program | |
Ko et al. | Internet Small Computer System Interface (iSCSI) Extensions for the Remote Direct Memory Access (RDMA) Specification | |
US20030074488A1 (en) | Method and apparatus for communicating between modules | |
US7392373B2 (en) | Granularity on a per port basis of request-response communications over a networked computer system | |
Ko et al. | RFC 7145: Internet Small Computer System Interface (iSCSI) Extensions for the Remote Direct Memory Access (RDMA) Specification | |
Stoess | Porting LibRIPC to iWARP |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A2 Designated state(s): AE AG AL AM AU AZ BA BB BG BR BY BZ CA CN CR CU CZ DM DZ EE GD GE GH GM HR HU ID IL IN IS JP KE KG KP KZ LC LK LR LS LT LV MA MD MG MK MN MW MX MZ NO NZ PL RO RU SD SG SI SK SL TJ TM TR TT TZ UA UG UZ VN YU ZA ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: A2 Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
WWE | Wipo information: entry into national phase |
Ref document number: 2000958987 Country of ref document: EP |
|
AK | Designated states |
Kind code of ref document: A3 Designated state(s): AE AG AL AM AU AZ BA BB BG BR BY BZ CA CN CR CU CZ DM DZ EE GD GE GH GM HR HU ID IL IN IS JP KE KG KP KZ LC LK LR LS LT LV MA MD MG MK MN MW MX MZ NO NZ PL RO RU SD SG SI SK SL TJ TM TR TT TZ UA UG UZ VN YU ZA ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: A3 Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE |
|
WWP | Wipo information: published in national office |
Ref document number: 2000958987 Country of ref document: EP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2387400 Country of ref document: CA |
|
NENP | Non-entry into the national phase |
Ref country code: JP |
|
WWG | Wipo information: grant in national office |
Ref document number: 2000958987 Country of ref document: EP |