Electrical computers and digital processing systems: multicomput – Computer network managing – Network resource allocating
Reexamination Certificate
1997-12-10
2004-04-06
Wiley, David (Department: 2142)
Electrical computers and digital processing systems: multicomput
Computer network managing
Network resource allocating
C709S241000, C709S225000, C370S432000, C370S397000
Reexamination Certificate
active
06718387
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention relates to computer networks and, more particularly, to the provision of highly-distributed servers for network applications, such as web service on the Internet.
2. Description of Related Art
Exponential growth in the number of clients and servers connected to large scale networks, such as the Internet, is causing increased problems of scale. At the same time that demand is accelerating, the network infrastructure is growing at a relatively slow rate. As these networks become more commercialized, the information and services provided by network servers represent a major revenue stream for many companies. The dependence of such companies on network services has led to increased demand for very high volume network servers and, particularly where companies have a global presence, there is a demand for distributed, uniform interface network servers that provide for graceful degradation in the face of problems in parts of the network. Components, such as host computers, packet switches, packet routers, gateways and the like, which are frequently used in construction of networks are well known in the art.
Many researchers have explored different mechanisms to improve the performance of network servers and systems. Some popular techniques include caching, and load balancing through the use of multiple servers. Some sites identify different servers with different names, each serving different geographic regions. Some approaches attempt to distribute access to a plurality of servers by using the domain name system (DNS) to randomize the server assigned when a connection request is directed to a particular company server. Some approaches attempt to utilize or mis-utilize the DNS by modifying the functionality of the DNS to poll each of the servers associated with a well known address to find out how loaded they are. The DNS then resolves the domain name to the IP address of the least loaded server.
The Problems
As demand for large network services has increased disproportionately with the underlying infrastructure to support the demand, the usefulness of such networks has been hampered by the congestion and bottlenecks which result. Currently, it is not uncommon for users to wait for tens of seconds and sometimes even minutes before they can get any information from the more popular (high-traffic) web servers. These delays frustrate the users and make them less likely to use a network for obtaining desired resources and services. This wasted time and effort represents a loss of productivity for network users and the resulting revenue losses are particularly undesirable for commercial Internet sites.
Typically, it would be desirable that a solution to the server problem have the following properties:
1. The provider should be able to set up many different web servers at locations all over a network such as the Internet, without any restrictions (such as requiring all web servers to be on the same subnet).
2. The clients (e.g. network users and their browsers) should be able to send requests to a single, well-advertised IP address.
3. Network servers should be able to, in coordination, choose/dictate the clients that they are willing to serve. They should not be required to listen to all the traffic from all clients. This assignment/selection may change with time and such changes should not impose excessive additional burdens on the routing system.
4. There should not be a single point of failure. Rather, within a reasonable recovery time period, all requests should be directed to any remaining server(s), although requests might experience longer service times (graceful service degradation).
SUMMARY OF THE INVENTION
In accordance with the invention, multiple distributed servers can be provided which accomplish load balancing and graceful degradation in the event of multiple network failures. The load is distributed in accordance with a load balancing policy which may range from the very simple to the very sophisticated.
Three basic approaches may be used individually or in combination. The first involves an extension of multicasting which can be characterized as “manycasting.” With manycasting, one can create highly distributed clusters that provide network services with all of the desirable properties described above. In manycasting, source specific “joins” and “leaves” may be utilized to establish the portion of the address space to be serviced by a particular server. Manycasting can be implemented in a secure form to prevent interlopers from intercepting or interfering with the communications. Using the first approach, four different techniques are utilized for reallocating connections when load redistribution occurs. They are (1) connection reset, (2) state synchronization and update, (3) message forwarding and (4) tag switching.
A second approach to overcoming the problems of the prior art involves an extension to the TCP protocol to enable dynamic TCP designations. This approach utilizes an extension to the SYN packet used in a TCP. With this option, the sender provides a tag and a cookie which the receiver can use. If the server is similarly equipped, it replies with a message that includes a tag, a cookie and destination information. When a server gets overloaded, it forwards all new connection requests to a less loaded server in accordance with the distribution policy. However, the same approach can be utilized to redirect a connection from an existing server to a different server. A security mechanism can be utilized to prevent the connection from being hijacked when a “change destination” message is sent.
The third major approach to solving the problems of the prior art utilizes tag switching. A pool of servers is supported behind at least one virtual IP address; virtual IP routers direct the packets for the virtual IP addresses to the server pool. The servers set up a family of tag switch trees (one for each real server). When a virtual IP router receives a tag-less packet, it forwards the packet to the actual IP address of a selected server and informs one or more upstream routers about the actual IP address to which subsequent packets should be directed. The upstream routers then mark all packets for that connection with a tag ID for the designated server. Thus, all subsequent packets will be correctly forwarded via tag switching.
The advantages of the present invention will become readily apparent to those skilled in the art from the following detailed description, wherein only the preferred embodiments of the invention are shown and described, simply by way of illustration of the best mode contemplated of carrying out the invention. As will be realized, the invention is capable of other and different embodiments, and its several details are capable of modifications in various obvious respects, all without departing from the invention. Accordingly, the drawings and description are to be regarded as illustrative in nature, and not as restrictive.
REFERENCES:
patent: 5187787 (1993-02-01), Skeen et al.
patent: 5321693 (1994-06-01), Perlman
patent: 5428615 (1995-06-01), Backes et al.
patent: 5557798 (1996-09-01), Skeen et al.
patent: 5774660 (1998-06-01), Brendel et al.
patent: 5835723 (1998-11-01), Andrews et al.
patent: 5835725 (1998-11-01), Chiang et al.
patent: 5881238 (1999-03-01), Aman et al.
patent: 5892924 (1999-04-01), Lyon et al.
patent: 5920705 (1999-07-01), Lyon et al.
patent: 5951694 (1999-09-01), Choquier et al.
patent: 5974453 (1999-10-01), Andersen et al.
patent: 5996021 (1999-11-01), Civanlar et al.
patent: 6006334 (1999-12-01), Nguyen et al.
patent: 6009474 (1999-12-01), Lu et al.
patent: 6018771 (2000-01-01), Hayden
patent: 6041343 (2000-03-01), Nguyen et al.
patent: 6061349 (2000-05-01), Coile et al.
patent: 6070191 (2000-05-01), Narendran et al.
patent: 6085238 (2000-07-01), Yuasa et al.
patent: 6128279 (2000-10-01), O'Neil et al.
patent: 6128657 (2000-10-01), Okanoya et al.
patent: 6188671 (2001-02-01), Chase et al.
patent: 6249820 (2001-06-01), Dobbins et al.
patent: 6272523 (200
Gupta Amit
Rom Raphael
Rosenthal & Osha L.L.P.
Sun Microsystems Inc.
Vaughn, Jr. William C.
LandOfFree
Reallocating address spaces of a plurality of servers using... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Reallocating address spaces of a plurality of servers using..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Reallocating address spaces of a plurality of servers using... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3233369