Chapter 9: Clustering Enhancements

Overview

Don’t tell my local bookstore, but I don’t shop there anymore-even though I’m frequently seen browsing the shelves. Instead, I browse the latest titles while sitting in one of the comfortable chairs the bookstore generously provides its customers (big mistake on their part) and when I find a book that interests me, I make a note of the title, author, and ISBN.

Then, when I get home, I order the book from an online bookstore. Shhh-don’t tell my local bookstore I do this, otherwise they might bar me from using their comfy chairs next time I visit them.

Online bookstores and similar sites have changed the way I do much of my shopping. But what if an online bookstore ran their entire Web site on a single server and that server died? Chaos! Frustration!! Lost business!!! I might even go back to my local bookstore and buy from there!

What keeps sites like these always available is clustering. A single server is a single point of failure for your business, and when that server goes down so does your revenue. The same goes for a single source of storage, a single network link, or even having all your computing resources located at a single geographical site. Fault-tolerant technologies such as RAID can mitigate the risk of storage failures, while redundant network links can reduce the impact of a network failure. And data backup and archival solutions are essential if you want to ensure continuity of your business after a catastrophe. But it’s also important to implement clustering technologies if you want to fully protect your business against downtime and ensure high availability to customers.

A cluster is simply a collection of nodes (servers) that work together in some fashion to ensure high availability for your applications. Clusters also provide scalability for applications because they enable you to bring additional nodes into your cluster when needed to support increased demand. And since the days of Microsoft Windows NT 4.0, there have been two types of clustering technologies supported by Microsoft Windows: server clusters and Network Load Balancing (NLB).

First, let’s look at server clusters. Originally code-named “Wolfpack” when the technology was first developed, server clusters provide failover support for long-running applications and other network services, such as file, print, database, or messaging services. Server clusters ensure high availability for these services because when one node in your cluster dies, other nodes take over and assume the workload of the failed node and continue servicing client requests to keep your applications running. In Windows NT 4.0, server clusters were known as the Microsoft Cluster Services (MSCS); in Windows 2000 Server, this feature was renamed Server Clusters. Now in Windows Server 2008, we call this technology Windows Server Failover Clustering (WSFC) or simply Failover Clustering, which communicates clearly the purpose of this form of clustering and how it works.

Then there’s Network Load Balancing, which was originally called Windows Load Balancing Service (WLBS) in Windows NT 4.0. This form of clustering technology was renamed Network Load Balancing (NLB) in Windows Server 2003, which is still the name for this technology in Windows Server 2008. NLB provides a highly available and scalable environment for TCP/IP services and applications by distributing client connections across multiple servers. Another way of saying this is that NLB is a network driver that balances the load for networked client/server applications by distributing client connections across a set of servers. NLB is especially great for scaling out stateless applications running on Web servers when the number of clients is growing, but you can also use it to ensure the availability of terminal servers, media servers, and even VPN servers.

Let’s look at the improvements the Windows Server team has made to these two clustering technologies in Windows Server 2008. As with everything in this book, the new features and enhancements I’m going to describe here are subject to change before RTM. And who knows? Maybe after you read this you’ll want to go out, buy Windows Server 2008, and start your own online bookstore! Well, maybe not-the competition is already pretty stiff in that market.