The request batching process helps to improve performance and reduce network traffic in two ways. First, fewer Web service calls occur between the client and the SharePoint server, which reduces the "chattiness" of the client-server interface. For example, you can perform two list queries in a single request. Second, as a set of operations occur on the server in a single request, the data being acted on doesn't need to be moved between the client and the server for the intermediate operations—only the list of operations and the final result set are passed between the client and the server.
Request batching requires a different mindset when you create queries from client-side code. First, be aware that you do not have access to any results until you call ExecuteQueryAsync (or ExecuteQuery) and receive the call back with the results. If you need to implement conditional logic in the client-side code that can't be expressed in the command list that you send to the server, you will need to execute multiple queries. Second, you should aim to group your operations to minimize the number of service calls. This means you may need to think about how you sequence your logic in order to take full advantage of request batching.
For more information about the client object model, see patterns & practices: Using the Client Object Model, patterns & practices: Client Application Models in SharePoint 2010, and SharePoint SDK: Using the Client APIs.