Electrical computers and digital processing systems: multicomput – Distributed data processing – Client/server
Reexamination Certificate
1998-09-22
2001-06-26
Harrell, Robert B. (Department: 2152)
Electrical computers and digital processing systems: multicomput
Distributed data processing
Client/server
C709S202000, C709S227000
Reexamination Certificate
active
06253230
ABSTRACT:
BACKGROUND
The Internet Engineering Task Force (IETF) Internetworking over NBMA (non-broadcast multiple access) (ION) working group is currently studying three different proposals for Internet Protocol (IP) Switching. These architectures can be summarized by two methods: an “Ipsilon” switching method associates Asynchronous Transfer Mode (ATM) connections to Internet Protocol flows; and a second that associates connections to egress routers routes.
ATM is well known in the art. By way of overview, ATM has its history based in the development of Broadband Integrated Services Digital Network (B-ISDN). ATM is a method of multiplexing and switching packets that has been chosen as the transmission mode of B-ISDN. ATM, a transfer mode for high-speed digital transmission uses a packet switching technology and has nothing to do with “asynchronous” transmission (see e.g., “PDH, Broadband ISDN, ATM and All That: A guide to Modern WAN Networking, and How It Evolved,” by Paul Reilly, Silicon Graphics Inc. (Apr. 4, 1994), which is hereby incorporated herein by reference in its entirety. ATM packets are called cells, wherein each cell has a 5 byte header and 48 bytes of data. ATM packet switching differs from conventional packet switching in that ATM packets follow pre-established routes called virtual paths and virtual circuits. Although ATM is not dependent on any particular physical medium of transmission, when the medium of transmission is mainly optical fibers, the error and loss rate is very small and hence no retransmission is done. See e.g., “Asynchronous Transfer Mode Tutorial,” Northern Telecom, http:/www.webproforum.com
ortel2/index.html, (Jun. 10, 1998), which is hereby incorporated herein by reference in its entirety.
The Transmission Control Protocol/Internet Protocol (TCP/IP) and the use of TCP/IP over ATM is also well known in the art. See e.g., D. E. Comer,
Internetworking with TCP/IP: Principles, Protocols, and Architecture,
Prentice Hall, Englewood Cliffs, N.J., (1988), which is hereby incorporated herein by reference in its entirety. Although Transmission Control Protocol (TCP) switching can work by allocating connections between the different ATM routers, e.g., using predefined Virtual Path Indicator/Virtual Channel Indicators (VPI/VCI), this method requires a given amount of packets to be exchanged per connection to be efficient. Another possible way to process is to use the Ipsilon IP switching method.
The traffic on the World Wide Web (Web) is increasing exponentially, especially at popular (hot) sites. Thus it is important to provide a scaleable web server (see for example, Goldszmidt, G. and Hunt, G. “Net Dispatcher a TCP Connection Router” IBM Research Report, 1997; and Dias, D. M., Kish, W., Mukheijee, R., and Tewari, R., “A Scalable and Highly Available Web Server”, Proc. 41st IEEE Computer Society Intl. Conf. (COMPCON) 1996, Technologies for the Information Superhighway, pp. 85-92, February 1996. One known method to provide load balancing in a scaleable web server is to use a so-called Network Dispatcher [see e.g., U.S. Pat No. 5,371,852, issued Dec. 6, 1994 to Attanasio et al., entitled “Method and Apparatus for Making a Cluster of Computers Appear as a Single Host,” which is hereby incorporated herein by reference in its entirety; and Attanasio, Clement R. and Smith, Stephen E., “A Virtual Multi-Processor Implemented by an Encapsulated Cluster of Loosely Coupled Computers”, IBM Research Report RC 18442, (1992). Here, only the address of the Network Dispatcher (ND) is given out to clients; and the Network Dispatcher distributes incoming requests among the nodes in the cluster (also called a virtual encapsulated cluster (VEC)), either in a round-robin manner, or based on the load on the nodes. In co-pending U.S. patent application Ser. No. 08/861,749, filed May 22, 1997, entitled “A Method for Local and Geographically Distributed Load Balancing Using A Generalized TCP Router”, by Dias et al., which is hereby incorporated herein by reference in its entirety, an example of a generalized Network Dispatcher is disclosed, that allows routing to nodes that may be located anywhere in a general inter-network.
The Internet backbone network is currently migrating to a switched ATM infrastructure. At the same time, very large servers (regardless whether they are a Mainframe, Mainframe clusters or other type of clusters) are being connected to the backbone via ATM links, to handle the dramatic growth in bandwidth and demands on throughput
In that context, the IETF is considering various alternatives to take advantage of the simple/fast/efficient processing capabilities of ATM switches. What the various alternatives have in common is a dynamic scheme to simplify all intermediate hops (any hop but Client and Server) processing by replacing a routing decision based on an IP header, with a switching decision based on an ATM header. This means that, ultimately, only the endpoints (i.e., Clients and Servers) will process the IP packets (IP layer, TCP layer, etc.) while any other hop on the path between endpoints will switch ATM packets. Some of the alternatives are also considering a so-called “short-cut” method, which is a mechanism to bypass some of the intermediate hops, when physical connectivity allows it. The solutions considered by the Internet community include: the Next Hop Resolution Protocol (NHRP) (see e.g., “Next Hop Resolution Protocol (NHRP)”, The Internet Society, Network Working Group, RFC 2332 (1998), which is hereby incorporated by reference in its entirety); the IPsilon IP Switching Protocols (IFMP and GSMP); Tag Switching; and IBM's Aggregate Route-based IP Switch (ARIS).
In a full or even partially switched network, a hop running a conventional front-end to a cluster of servers (such as the Network Dispatcher) would conflict with the entire innovative approach being studied by the IETF. It would have to examine the IP and TCP fields, while any other hop would be trying to avoid considering IP to make a routing decision.
SUMMARY
In accordance with the aforementioned needs, the present invention has features which provide a switching capability in a front-end to a cluster of servers so that packets can be switched up to the Server, and back to the client, or to the closest Switch to the client.
One version of the present invention includes two separate components: a Control Engine (CE); and a Forwarding Engine (FE). The Control Engine is in charge of assigning a connection to a server and forwarding information about the assigned server and connection to the Forwarding Engine. Each Forwarding Engine applies the assignments received from the Control Engine to map a TCP connection onto a switched ATM connection. At the end, the Forwarding Engine communicates the connection termination event back to the Control Engine.
An example of a method for selecting a server from a cluster of servers and a switched path to a selected server in a client-server system including a switched network, includes the steps of: a forwarding engine (FE) receiving a client request; the FE routing a request to a control engine (CE) to select a server from the cluster and a corresponding switched address, in response to the client request; the CE selecting a server and communicating the corresponding switched address to the FE; and the FE forwarding data associated with the client request to a selected server over a switched connection associated with the switched address, wherein the switched connection need not traverse the CE.
The step of the CE communicating the switched address to the FE may further include the step of, the CE communicating server selection criteria to the FE and conditions under which the FE may use the criteria; and for subsequent client requests received by the FE:
the FE determining if there is an existing switched connection associated with this request; and
if there is an existing connection, the FE forwarding the request over the existing switched connection; and
if there is no existing switched connection, the FE selecting the destination serv
Couland Ghislaine
Hunt Guerney Douglass Holloway
Jean-Marie Mauduit Daniel Georges
Levy-Abegnoli Eric Michel
Harrell Robert B.
International Business Machines - Corporation
Scully Scott Murphy & Presser
Thompson Marc D.
Zarik, Esq. Gail H.
LandOfFree
Distributed scalable device for selecting a server from a... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Distributed scalable device for selecting a server from a..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Distributed scalable device for selecting a server from a... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2490506