[Top] [Prev] [Next] [Bottom] [Contents]

Deploying Sapphire/Web Projects

4

Introduction

A Web Application can be broken down into a number of different components:

Browser and Related Software
Internet / Intranet Transmission Software
Web Server
Server-side Application Software
Database Access Software
Database Management System and Data

You use Sapphire/Web to create the Application Software portion of this structure. Sapphire/Web can produce a number of different types of application software.

CGI - a program that is run by the Web Server in order to service a request originating at the browser and which does its work and then exits completely
Application Server: a program that can be launched to service a browser request, as above. However, this program does not exit on completion of the task. This behavior eliminates the time and CPU costs associated with continually relaunching instances of a CGI application.
- Sapphire/Web Application Servers
- Oracle Web Cartridges

Application Server Features

Application Servers have the advantage that they can run on another machine or machines, rather than the Web Server's machine
Load can be spread automatically between machines using application servers. Instances of an application server program are only launched as they are required, and will terminate automatically - either after a certain usage count or after a certain amount of idle time - thus releasing resources.

For Sapphire/Web Application Servers the follow points apply:

Distribution. Each "App. Srvr." in the diagrams below is representative of a different application program, perhaps on a different physical machine - possibly even on a different type of machine and operating system.
Instances. Each application server program is defined in the config file. It is defined as being a certain (.exe) file at a certain location (in a machine's file system) for loading purposes. Also defined there are the number of instances of the same program that can be run simultaneously on the defined machine.
Load Balancing. One can control the load imposed on a machine by an application server by controlling the number of simultaneous instances of a program one permits. Application Servers created by Sapphire/Web can be controlled in that they can be limited to a certain number of simultaneous instances.
Fire Walls. One advantage of using an application server is to be able to have the Web Server outside the fire wall but still be able to deploy accessible applications that remain inside the fire wall. This is shown in the second diagram Conversion

To convert your project from one type of executable to another is very simple. If you have created your project entirely within Sapphire/Web you need only alter a few settings in the Project Options dialog box and re-create your application. It will now be created in the new format.

Sapphire/Web comes with a set of extra programs and DLLs to allow you to create application servers in any of the categories listed above. These supply you with

An NSAPI plugin to perform the same function for Netscape servers
An Application Server Manager program to control the Application Servers built with Sapphire/Web
An Application Server Manager Console program to provide a user interface for the control functions implemented by the Application Server Manager
An Application Server CGI program to allow access to Application Server programs from web servers that do not support Netscape NSAPI

After this, only the step of "registration" remains. This step is necessary to inform your web server of the existence of the application server. This is done by modifying the Application Servers configuration file with the Application Server Manager Console program.

For an Oracle Web Cartridge the process is similar; you move the .DLL to the proper subdirectory and use the Oracle Web Server Administration application to tell it about this new Web Cartridge.

Session Affinity

If an application server is configured to use Session affinity than you can be sure that a user accessing your application will always contact the same application server during a session. This is useful if you want to keep some kind of state information in the memory of the application server.

Load Balancing

All Web Server Extensions (WSE) provide a form of load balancing to spread the load of users accessing an application. This is achieved by using two scheduling algorithms. The default algorithm is used when an application server is defined to not use session affinity. You can think of this algorithm as an on demand scheduler. When a WSE gets a request to connect to an application server it searches the application server configuration file for the first instance that is running and is not in use and connects that instance. If all running instances are busy a new instance will be started to handle the new request.

If all instances defined in the configuration file are in use and already started than it will connect to the instance with the least number of queued connections. The benefit of this algorithm is that application server instances are launched to handle the current load of users on the application.

Thus if you define five instances in the configuration file and five different users access this application at different times, only one application server will be started. The second scheduling algorithm is the similar to the one above except it ensures that a user of an application always connects to the same instance and always starts a new instance with a new user until the maximum number of instances defined in the configuration file are started. This is needed for session affinity. Note that this algorithm can cause more instances to be started on your server than the one above so make sure you only set the session affinity flag if you are storing information inside of an application server instance. You do not have to use session affinity if you are using the Sapphire/Web State Server.

[Top] [Prev] [Next] [Bottom] [Contents]

info@bluestone.com

Copyright © 1997, Bluestone. All rights reserved.