Electrical computers and digital processing systems: multicomput – Computer network managing
Reexamination Certificate
1996-07-01
2002-03-19
Harrison, Jessica J. (Department: 3713)
Electrical computers and digital processing systems: multicomput
Computer network managing
C709S241000, C709S203000, C709S207000, C709S218000, C709S227000, C709S249000
Reexamination Certificate
active
06360256
ABSTRACT:
FIELD OF THE INVENTION
This application relates to distributed computing using the client/server model and, more specifically, to an improved name service for a client/server system.
BACKGROUND OF THE INVENTION
In recent years, the internet has become extremely popular. Using the internet, a group of users in widely scattered locations can retrieve information located at a centralized site. The site, which may offer one or more “services” (also called “resources”), is connected to the internet through a high speed connection. A user can access the site via the internet to read electronic mail, read news via Usenet, or view World Wide Web pages. As the speed of the connections between the user and the site improves, more users will be attempting to access the site's resources at a given time.
Internet services in a site are accessed via the servers of the site. Multiple servers can provide access to a single service. Similarly, a single server can provide access to multiple services.
Resources on the internet are accessed via names. A resource name typically contains multiple name components. The first name component is a domain name. This domain name typically refers to one or more hosts and conventionally is resolved through the Domain Name Service (DNS) of the site. An implicit or explicit second name component identifies the communications endpoint or server that will be used to access the resource on a particular host. The remaining name components are processed by the identified server to access the specified resource on the located host.
For example, a WWW hypertext document is named by a Uniform Resource Locator (URL), such as:
http://www.sun.com/welcome.html.
To access this document, the domain name component, www.sun.com, is first resolved to locate a host. Then, the hypertext transfer protocol (http) name component identifies a communications endpoint on the located host. An http server communicates with the client through this endpoint and retrieves the document specified by the remaining component of the name, /welcome.html. The File Transfer Protocol (FTP) service uses similar URLs.
Usenet news articles are named in a similar way. To obtain a news article, the news reader client resolves a domain name to locate a news server. It then communicates with the news server using a well-known endpoint to retrieve a particular article on the news server.
Many internet systems include a process called a “name service” that examines the type of service required by each incoming request and returns a list of addresses of servers that can handle the request. It is the responsibility of the name service to balance the workload of the system by returning the addresses in such a way that work is spread evenly between the servers.
In a redundant server system, a site has more then one server. These servers collectively serve the entire user population. A problem arises when certain servers are performing more than their share of the workload of incoming user tasks. Some servers are operating at or near their capacity while other servers that could be sharing the workload remain idle. Such a system is said to have an “unbalanced” workload. It is desirable to balance the workload of the servers as much as possible.
In addition, servers go in and out of service. Because any server can serve any user, the workload of a failed server can be assigned to other working servers. A conventional redundant server system effects load balancing using a “round robin” scheme to assign incoming requests to servers. As each request is received, the DNS service assigns a next server of the available servers. A round robin scheme for load balancing is not always satisfactory because the name service does not always have up-to-date information about which servers are actually available. For example, a server may have failed or may have been added to the system without the name service being notified. Conventional name services often rely on a human being to reconfigure the knowledge that the name service has about the system.
A site installs one or more internet servers to gain access to the internet and its services. The primary goal is to enable a site to deploy its internet servers in a highly “scalable” and “reliable” manner. A “scalable deployment” architecture allows the site to support a large number of users and to increase capacity gracefully as the number of users increases. A “reliable” architecture minimizes the impact of component failures. It may mask failures by redistributing the workload of failed components to working components.
SUMMARY OF THE INVENTION
The present invention overcomes the problems and disadvantages of the prior art by implementing a self-reconfiguring name service that distributes workload among the available servers in a system. A Service Monitor for each host system of a site periodically broadcasts information about available servers. As new servers are added to the host, they are announced by way of this broadcast message. The broadcast message also indicates the workload of the host.
Each name service has an associated process called a Name Binder Modifier that receives the broadcast messages from the Service Monitors. Periodically, each Name Binder Modifier reviews the information it has received from various Service Monitors. The Name Binder Modifier maintains a list for each service of available servers. For each service, servers executing on a host whose workload exceeds a predetermined workload value are deleted from a list of available servers. The list of available servers for each service contains at least a minimum number of servers whenever possible, however.
The Name Binder Modifier periodically updates a plurality of zone files for respective services in accordance with its lists to indicate which servers are available to have work routed to them. The Name Binder Modifier assumes that servers which have not broadcast a message within an update interval are no longer available. The DNS service loads information from the updated zone files for use in its routing scheme.
Each Service Monitor has an associated configuration file specifying, for each service, a time interval after which the Service Monitor should send a broadcast message. Furthermore, each Name Binder Modifier has a configuration file specifying, for each service, a time interval after which the zone files should be updated in accordance with the broadcast messages, a minimum number of servers for the service, and a maximum desirable workload per server.
In accordance with the purpose of the invention, as embodied and broadly described herein, the invention is a method for reconfiguring a load balancing system, comprising the steps, performed by a data processing system, of sending, by a Service Monitor, for a service of the data processing system, a broadcast message containing a workload of a host on which the service is located and a list of addresses of the host; receiving, by a Name Binding Modifier, the broadcast message; and updating, by the Name Binding Modifier, a zone file in accordance with the information in the broadcast message.
Objects and advantages of the invention will be set forth in part in the description which follows and in part will be obvious from the description or may be learned by practice of the invention. The objects and advantages of the invention will be realized and attained by means of the elements and combinations particularly pointed out in the appended claims.
REFERENCES:
patent: 4642758 (1987-02-01), Teng
patent: 5410651 (1995-04-01), Sekizawa et al.
patent: 5539883 (1996-07-01), Allon et al.
patent: 5740371 (1998-04-01), Wallis
patent: 5774660 (1998-06-01), Brendel et al.
patent: 5774668 (1998-06-01), Choquier et al.
Dias et al. “A Scalable and Highly Available Web Server” pp 85 -92, IEEE, Feb. 1996.*
Zhu et al. “An Experimental Study of Load Balancing on Amoeba” pp 220 -226, IEEE, Mar. 1995.
Cherubin Yveste G.
Harrison Jessica J.
O'Melveny & Myers LLP
Sun Microsystems Inc.
LandOfFree
Name service for a redundant array of internet servers does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Name service for a redundant array of internet servers, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Name service for a redundant array of internet servers will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2838710