Before making any decisions about deploying your enterprise SharePoint globally, consider the priorities and preferences of your stakeholders.
And, as always, evaluate the cost versus usability for every deployment scenario. One of the early challenges when planning a global SharePoint deployment is determining how many farms are needed and where they should be located. The next challenge is understanding what might be lost by providing regional or distributed farms.
Keep in mind that cost shouldn't always dictate the SharePoint deployment model you choose. Finding the right answer for your deployment may take some testing.
That said, the three main distribution models for deploying Microsoft Office SharePoint Server (MOSS) 2007 are the following:
SharePoint deployment option 1: Centralized model
The centralized model is the utopia of SharePoint Server deployments from an operations and usability perspective because it consists of a single location (See Figure 1). Because there is just one site, users in a centralized deployment have a common operations team, a common service level agreement and a common site provisioning model for collaboration sites. Users might also prefer going to a single place for enterprise search.
Remote users across the globe can make a centralized deployment a little less desirable, however, with slow or saturated WAN links. A poor WAN link has more than 150 milliseconds of latency, but it can vary as bandwidth is increased and usage is decreased. Although latency is the greatest challenge, bandwidth and utilization do have a significant impact on speed of downloads.
There are a number of tools available to overcome the effects of latency. The most popular tools—which come from companies from Cisco to Citrix and even F5—are WAN accelerators and WAN acceleration devices. Common features found in these tools include compression, differencing and various proprietary network optimization techniques. IT managers will see these devices and their distinct network optimization advantages leveraged across platforms as they become more commonplace.
SharePoint deployment option 2: Regional model
Some networks are designed with a hub-and- spoke configuration (See Figure 2). Usually it's companies and organizations that drive IT shops to provide multiple data centers as network equipment becomes cheaper, leased lines become dedicated and fiber gets more common. Offices can't live without fast access to the Internet or without fast access to their Web applications. IT drives consolidation, but companies have to decide what to do with their data storage and consolidation strategies.
Companies that are distributed across the globe often find some "regional hub" as the answer to providing a consolidated solution for Microsoft Exchange, Active Directory or even line-of-business applications. Looking at these other solutions can provide an example—if not a model— for how to lay out your regional SharePoint deployment.
Crawling and indexing becomes the challenge with multiple WAN locations. With limited bandwidth between locations, finding the utilization to do recurring indexing without saturation issues can be accomplished with careful planning and by working closely with the network team doing monitoring. But as the data increases on either side of the WAN, the time it takes to crawl continues to increase.
Even with the incremental crawl and the crawl based on the change log—both of which really help to minimize crawl time—the most recent answer is federated search. First introduced with search server, federated search is now available in the infrastructure update and is sure to be included in Service Pack 2 of SharePoint Server.
SharePoint deployment option 3: Distributed model
If every office had a SharePoint deployment, then companies would have hundreds of SharePoint farms. Some businesses do dictate that each affiliate should have its own data that's separate from another. In businesses where speed is the key factor, replacing file servers with SharePoint servers is the first step to empowering the business. In reality, there may be a dozen reasons why businesses find it easier to simply provide SharePoint to each department or each affiliate (See Figure 3).
It's easy to do a simple deployment in each location, but a distributed SharePoint deployment model is the most difficult to support. By far, a distributed model is the most difficult to operate from a backup, maintenance, and storage allocation and optimization perspective.
What are the most important considerations when deciding on a topology for any SharePoint deployment? Know your network—especially its bandwidth, latency and utilization. Keeping these in mind, even those that start with a centralized deployment by default and move onto a regional or distributed deployment will find that their hard work does pay off.
|Joel Oleson is an independent consultant involved in training, speaking, technical evangelism and product management for a variety of companies including Nintex. As former senior technical product manager for Microsoft Office SharePoint Server, Oleson focused on topics related to enterprise deployments of SharePoint, such as performance, scale, backup/restore and high availability. His blog is SharePoint Joel's SharePoint Land.|