Load balancing of data is more efficient in DFS than in standard nondistributed file systems. One reason is the use of replication, which allows DCE LFS filesets to be replicated on multiple machines. Requests for files from frequently used DCE LFS filesets are then spread across different machines, preventing any one machine from becoming overburdened with data requests. Multihomed server capability makes it possible for each machine to have multiple connections to the network, allowing direct connections to the subnetworks that provide the most requests. These connections help reduce cross-router traffic within the network.
Fileset characteristics in DCE LFS also improve load balancing. DCE LFS filesets are typically smaller than standard UNIX and other non-LFS filesets; DCE LFS aggregates can accommodate multiple DCE LFS filesets for flexible disk usage; and DCE LFS filesets can be moved between aggregates on different File Server machines. The ability to store multiple filesets on a single aggregate is integral to being able to move filesets in DFS.
DFS automatically tracks every fileset's location, even when the fileset is moved between aggregates or machines. The location of any fileset is automatically maintained in DFS by the Fileset Location Database (FLDB). This database tracks the machine and aggregate that houses each exported fileset (DCE LFS or non-LFS). Therefore, the user never needs to know the machine or aggregate that actually houses the fileset.
The FLDB relieves the system administrator of the burden of manually tracking each fileset's location, thus freeing the administrator to concentrate on more important administrative duties. Also, the master version of the database is typically replicated and synchronized (using Ubik) on multiple server machines, making access to the FLDB more reliable.