Administering an ISP Installation |
Now it’s time to quickly review the growth of domain.com. In this sample scenario, you started with a single IIS 5.0 Web server, from which you essentially ran everything. Since the customer’s application demanded highly available, yet scalable, performance, you built a three-tier Web architecture and implemented high-availability Microsoft technologies to each tier. You then tested the single server with the Web Application Stress Tool and realized that, to achieve even higher availability, you needed to build a Web cluster.
To do this, you initially calculated that three servers would suffice. Then you installed Network Load Balancing, whereby three servers essentially appeared as one. Next, you added a fourth Web server as the development or staging server, to which your customer could publish. In order to replicate the developed content, you implemented the Content Deployment feature of Site Server 3.0 on the staging server and the production Web servers. The customer could then connect through FrontPage to the staging server (domain.com), and you were able to easily replicate content to the production Web servers.
To increase performance in the second tier, you added a COM+ application and used ASP applications from the first tier to call on precompiled components from the second tier. These second-tier components called on the third-tier data service.
A high-end SMP server running SQL 7.0 made up the third tier of the system. To achieve high availability, you added Cluster Service, to ensure a fault-tolerant failover system. This topology provides high availability and scalability in spite of planned failures (such as service-pack or hardware upgrades) or unplanned failures (such as hardware failure or the loss of software integrity). If you are an ISP, you can deploy this architecture for your enterprise-level customers who run e-commerce sites or other mission-critical business applications on the Web. You can also use this topology for your common or commoditytype hosting, in which many customers participate within the single cluster.