REDMOND, Wash., Feb. 12, 2001 — On the back end, there are databases, file servers and messaging mechanisms — the source of goods and services to be delivered. On the front end, there are users or the consumers of the goods and services, and their PCs or other computing devices. Servers connect the two and facilitate interactions and exchanges.
Historically, big companies have increased their capacity by purchasing a big, expensive server — known as
“big iron”
— while smaller businesses have made do with what they could afford until they were able to expand. But recently that began to change as businesses adopted the use of clusters of smaller servers.
The only problem? A system administrator had to administer every server in the cluster separately.
Today, Microsoft announced the release to manufacturing (RTM) of Application Center 2000, a deployment and management tool for businesses that are using clusters of servers instead of taking the
“big-iron”
approach. By clustering servers, Microsoft says, businesses can enjoy higher availability, easier manageability and greater scalability then ever before. And now, using Application Center 2000, the clusters of servers can be managed with the same — or better — ease and cost effectiveness as managing a single server.
“This is the missing piece in the puzzle of clustering servers,”
says Bob Pulliam, Microsofts technical product manager for Application Center 2000.
Pulliam says the capabilities offered in Application Center 2000 provide a compelling alternative to big-iron servers.
“Businesses that buy a big-iron solution are buying fixed capacity,”
he says, using a company that runs a tax application as an example.
“If you were to buy a big-iron solution youd have to buy a box that was big enough to handle the first two weeks of April and use it year-round — regardless of the demand,”
he says.
“With the scalability thats available with clustering, you can adjust how many servers handle certain applications based on what the demand is.”
That capability, Pulliam says, is critical to the success of an Internet-based business.
“Say Im running a Web site that gets 10,000 hits per day,”
he says.
“If that shoots up to a million hits per day because of a seasonal factor or a sudden jump in popularity, I need to make sure that my servers can handle a dramatic increase in volume. Theres nothing worse for an Internet-dependent business than a URL thats down.”
Developed with Customers In Mind
Pulliam says that Microsoft looked to customer input for guidance when developing Application Center 2000.
“We had focus groups to determine how people were using servers,”
he says. As is the case across the whole Microsoft .NET Enterprise Server family, one of the criteria for shipping Application Center 2000 was to have customers sign off on the products readiness.
One such customer for Application Center 2000 is Dallas-based Data Return Corporation. Data Return provides managed hosting services to enterprise clients that utilize Microsoft technologies for their Web-based applications, including RadioShack.com, H & R Block, Texas Instruments and Pier 1. Data Return began using Application Center 2000 on its extranet site (www.datareturn.com) in November 2000.
“Managing applications across a Web cluster has always been complicated and often problematic,”
says Jason A. Lochhead, co-founder and chief technology officer of Data Return Corp.
“Application Center 2000 is the first product I have seen that really addresses this issue.”
Lochhead says that Application Center 2000 has a number of time- and cost-saving features, and that it’s simple and fast to deploy.
“Creating an Application Center 2000 cluster is very easy,”
he says.
“Just install Application Center on the servers, add them to the cluster, and synchronize. Once you are up and running, changes to your application can be synchronized across the Web cluster with a single click, or replicated automatically.”
Application Center 2000 makes considerable cost savings possible, according to Lochhead.
“By automating most of the tasks required to deploy or update an application, Application Center 2000 can help reduce the cost of labor associated with maintaining your Web cluster,”
he says.
Scaling the Middle Tier
Web servers exist in a zone between the front and back ends that is known as the
“middle tier.”
While the middle tier is typically invisible to either end, its failure to function properly — and instantly — can be disastrous. Like the heart and soul of any infrastructure, the importance of the middle tier is often underestimated until it malfunctions.
And by then, according to Pulliam, its usually too late.
“Historically, there have been problems associated with scaling the middle tier,”
Pulliam says.
“If the middle tier cant rise to accommodate a sudden and unexpected rise in demand, that can, and often does, mean big trouble for a business. For that reason, making the middle tier as scalable as possible became our main priority during the development of Application Center 2000.”
On the long road to providing server clusters with higher scalability, according to Pulliam, three key areas emerged. Server clusters need to be easily managed, the load on the network needs to be automatically balanced among the servers in a cluster and the clusters must have built-in and automatic maintenance mechanisms to ensure that the system is always available.
Application Center 2000s synchronization features, Pulliam says, allow customers to manage all the servers in the cluster like a single server.
“Servers across the cluster need to be configured the same in order to run the applications in a way that is consistent,”
he says.
“Otherwise youd have different servers in the cluster receive the same identical information and interpret it in two different ways.”
Application Center, according to Pulliam, keeps content and configuration settings synchronized across the cluster, providing the foundation for easier and less complex server management.
In describing the Network Load Balancing (NLB), Pulliam likens it to a card dealer who distributes a deck of playing cards to each player — NLB keeps the distribution of data and applications appropriately distributed across the cluster.
“This is at the core of increasing capacity through scaling out, which spreads the load across multiple servers and therefore eliminates a single point of failure, rather than scaling up, which means you pull up a truck and unload a larger server to replace your old one.”
Finally, Microsoft focused on developing maintenance mechanisms to monitor the site and keep availability as high as possible.
“In the past, people on inline skates in the data center would go from server to server to make sure they were healthy,”
Pulliam said.
“Now, they have the ability to look at performance and event and health information from a single console.”
Application Center 2000, Pulliam says, takes it one step further.
“Its one thing to look at a screen and see a red light that warns about memory or a network problem,”
he says.
“We give users the ability to configure a server to look for problems. You can configure a server so that if memory gets low it automatically pulls itself out of load balancing, fixes itself, restarts the application and comes back into the load balancing loop.”
On-Demand Scalability
Imagine an online shopping site. Its the day after Thanksgiving, and the server, overwhelmed by the sudden jump in activity, goes down. For businesses operating on one server, when that server gets overwhelmed and breaks down it means the URL is down also — an embarrassment , and a possible financial nightmare.
Hopefully the breakdown wouldnt occur on a high-profile shopping day. But the problem is this: A business has no control over when its big iron server crashes.
Application Center 2000 also makes it easier to scale the cluster on-demand, Pulliam explains, which gives customers new agility, enabling them to quickly respond to changes in their business needs.
“We have a wizard that will guide customers through a few simple steps to add a new server to the cluster. Application Center takes care of everything else. The copying, configuring, all of the things that have to happen in order for the server to be ready for action — we developed Application Center 2000 to do that automatically.”
Adding a server for one customer, Pulliam reports, went from a task that took at least two hours to one that could be completed in less than 20 minutes.
“Application Center 2000 spreads the business out among several smaller servers instead of relying on one large one, and thereby eliminates a single point of failure and increases the availability of the applications,”
Pulliam says.
