Method and apparatus for making a cluster of computers appear as a single host on a network The present invention provides a method and apparatus for enabling a cluster of computers to appear as a single computer to host computers outside the cluster. A host computer communicates only with a gateway to access destination nodes and processes within the cluster. The gateway has at least one message switch which processes incoming and outgoing port type messages crossing the cluster boundary. This processing comprises examining certain information on the message headers and then changing some of this header information either to route an incoming message to the proper computer node, port and process or to make an outgoing message appear as if originated at the gateway node. The message switch uses a table to match incoming messages to a particular routing function which can be run to perform the changes necessary to correctly route different kinds of messages. Continuously available database server having multiple groups of nodes, each group maintaining a database copy with fragments stored on multiple nodes A database server with a "shared nothing" system architecture has multiple nodes, each having its own central processing unit, primary and secondary memory for storing database tables and other data structures, and communication channels for communication with other ones of the nodes. The nodes are divided into first and second groups that share no resources. Each database table in the system is divided into fragments distributed for storage purposes over all the nodes in the system. To ensure continued data availability after a node failure, a "primary replica" and a "standby replica" of each fragment are each stored on nodes in different ones of the first and second groups. Database transactions are performed using the primary fragment replicas, and the standby replicas are updated using transaction log records. Every node of the system includes a data dictionary that stores information indicating where each primary and standby fragment replica is stored. The records of each database table are allocated as evenly as possible among the table fragments. A transaction manager on each node responds to database queries by determining which fragment of a database is being accessed by the query and then forwarding the database query to the node processor on which the primary replica of that fragment is stored. Upon failure of any one of the data processors in the system, each node updates the information in its data dictionary accordingly. In addition, the fragment replicas made unavailable by the node failure are regenerated and stored on the remaining available nodes in the same node group as the failed node. Interprocess communications interface for managing transaction requests The present invention provides a method of processing transaction requests from client applications within a computer network having a plurality of client servers. Each client server has a work share and a set of attributes that including a name, an address and a list of services for fulfilling transaction requests. Each transaction request identifies attributes necessary for fulfilling the request. The method entails identifying a set of client servers having the necessary attributes and defining a work distribution function. The work distribution function, which distributes transaction requests, randomly selects a client server from the set of client servers, according to work shares of the client servers. World-wide-web server with delayed resource-binding for resource-based load balancing on a distributed resource multi-node network A multi-node server transmits world-wide-web pages to network-based browser clients. A load balancer receives all requests from clients because they use a virtual address for the entire site. The load balancer makes a connection with the client and waits for the URL from the client. The URL specifies the requested resource. The load balancer waits to perform load balancing until after the location of the requested resource is known. The connection and URL request are passed from the load balancer to a second node having the requested resource. The load balancer re-plays the initial connection packet sequence to the second node, but modifies the address to that for the second node. The network software is modified to generate the physical network address of the second node, but then changes the destination address back to the virtual address. The second node transmits the requested resource directly to the client, with the virtual address as its source. Since all requests are first received by the load balancer which determines the physical location of the requested resource, nodes may contain different resources. The entire contents of the web site is not mirrored onto all nodes. Network bottlenecks are avoided since the nodes transmit the large files back to the client directly, bypassing the load balancer. Client browsers can cache the virtual address, even though different nodes with different physical addresses service requests. Dynamic link page retargeting using page headers A method for distributing the workload among multiple replicated servers in a computer network, preferably an Internet environment utilizing web pages and web servers, whereby the only changes are to the process or the code reside on the web server and there is no reduction in availability of the resources.