You replicate DCE LFS filesets by placing read-only copies of them on one or more File Server machines in a cell. If a machine that houses a read-only copy of the fileset becomes unavailable, the information is usually still available from a copy of the fileset on another machine. However, for a fileset that uses Release Replication, if the read-only fileset that resides at the same site as the read/write fileset becomes unavailable, all other read-only versions of that fileset become unavailable after a configurable amount of time. Similarly, for a fileset that uses Scheduled Replication, if the read-write fileset becomes unavailable, all read-only versions of the fileset become unavailable after a configurable amount of time. (See Making Filesets and Aggregates Available for detailed information on the availability of read-only filesets.)
Replicate only those DCE LFS filesets that meet the following criteria:
· The files in the fileset are read much more often than they are modified.
· The files in the fileset are heavily used; for example, binary files for text editors or other popular application programs. Replicating the fileset lets you distribute the load for the files that it contains across several machines.
· The files in the fileset must remain available. By replicating the fileset on multiple File Server machines, even if one of the machines that houses a replica of the fileset becomes unavailable, replicas are usually still available from other machines.
· The fileset is mounted at a high level in the cell's file tree; for example, root.dfs and its subdirectories.
If your cell is large, you may want to use a small set of File Server machines to store just read-only filesets. These machines can then distribute frequently used data, lessening the load on other machines. Keep in mind that each replica not stored on the same aggregate as its read/write source fileset uses as much disk space as its source fileset. A read-only fileset created on the same aggregate as its source fileset is created as a clone of its source and so requires potentially much less space than a full read-only replica created on a different aggregate.
Each Cache Manager maintains preferences in the form of numerical ranks that bias its selection of File Server machines for read-only fileset access. When accessing a read-only fileset, the Cache Manager consults its collection of preferences and attempts to access the read-only fileset from the File Server machine that has the lowest recorded rank. If the Cache Manager cannot access the fileset from that machine, it tries to access the fileset from the machine that has the next lowest rank. It continues in this manner until it either succeeds in accessing the fileset or determines that all of the machines that house the fileset are unavailable.
By default, the Cache Manager assigns preferences to File Server machines based on IP addresses. You can set or change the Cache Manager's preferences to suit your needs. (See Configuring the Cache Manager for more information about the Cache Manager and File Server machine preferences.)