A computing system's reliability and availability are the measure of its ability to provide desired services when required. Reliability and availability are critical requirements in modern computing systems that must be available 24 hours a day, 365 days a year.

Scalability is the ability of a system to maintain a nearly linear response time with increasing simultaneous users. If the processing load increases, we can maintain the performance of the application by scaling the servers. We can either add more servers to the cluster or add more CPUs to the servers in the cluster. We can scale the entire system without making the application unavailable, by upgrading one server at a time.

The COSS Internet system includes the following technologies to provide reliability, availability, and scalability:

Multi-Tier Architecture
In the COSS Internet system, software components are grouped into multi-tier logical architecture and deployed into a multi-tier physical architecture. As a Windows DNA-based application, the COSS Internet system multi-tier architecture is a fundamental requirement for building a scalable application. Instead of designing an application to run solely on one computer, it is designed to run on multiple tiers of multiple computers.

Assuming the next application is the illustration system, the client and producer data in the XMLife file are imported and the data is displayed automatically in the illustration screens. After the agent successfully sells a product to the client, the illustration application creates a new XMLife file containing the client, producer, and new data such as products and benefit information. The new XMLife file is available for transfer and use by another COSS application, such as the Advanced Marketing or iForms module. At any point during the sales process, information can be transferred out of the COSS system into a third-party application capable of reading XMLife files.

As shown in the architecture, the system architecture has three physical tiers. The Presentation Tier includes a Web server farm that uses Windows 2000 Advanced Server running with Network Load Balancing (NLB) and the Internet Information Server (IIS) version 5.0. The Business Logic Tier includes an application server farm that use Microsoft Application Center 2000 server running COM+ service and load balancing. The Data Tier includes a pair of clustered database servers. These database servers are configured to monitor each other's heartbeat, and implement a "fail over" plan if one fails. Database servers share mirror disk arrays to further ensure the data services of the application.

Load Balancing
Load balancing is a solution to distribute HTTP traffic to a group of Web servers. The load balancing enables the outside world to look at the entire group of servers as a single server. Load balancing provides many benefits. One is to make the system scalable, reliable, and available. The other is to help IT staff maintain and upgrade hardware and software more easily. There are two typical solutions for load balancing to a Web server farm: a) Round Robin DNS (hardware) and b) Intelligence Traffic Management (software). CISCO provides a Round Robin DNS device named "Local Director" as a hardware solution. Microsoft's Network Load Balancing (NLB) is the software solution.

On the Web server farm, COSS use Network Load Balancing to distribute HTTP requests to the system. NLB has a built-in solution to handle application state. Application state refers to data maintained by a server application on behalf of its clients such as a client's session. Please note that Application Center 2000 servers in the Business Logic Tier also use the NLB to increase the availability of application components.

Clustering
Clustering is similar to load balancing. A cluster is a group of independent systems that operates as a single virtual system. Clustering is different from load balancing in that instead of distributing HTTP request traffics, it distributes the business process work to the same objects that reside in many different application servers. Clustering technology provides many important capabilities such as session management of users, transparent replication, load balancing, and fail over. Clustering solutions are usually applied at the business logic server farm and database servers.