Deduplication jobs

Deduplication is performed by a system maintenance job referred to as a deduplication job. You can monitor and control deduplication jobs as you would any other maintenance job on the cluster. Although the overall performance impact of deduplication is minimal, the deduplication job consumes 400 MB of memory per node.

When a deduplication job runs for the first time on a cluster, SmartDedupe samples blocks from each file and creates index entries for those blocks. If the index entries of two blocks match, SmartDedupe scans the blocks adjacent to the matching pair and then deduplicates all duplicate blocks. After a deduplication job samples a file once, new deduplication jobs will not sample the file again until the file is modified.

The first deduplication job that you run might take significantly longer to complete than subsequent deduplication jobs. The first deduplication job must scan all files under the specified directories to generate the initial index. If subsequent deduplication jobs take a long time to complete, this most likely indicates that a large amount of data is being deduplicated. However, it can also indicate that users are storing large amounts of new data on the cluster. If a deduplication job is interrupted during the deduplication process, the job will automatically restart the scanning process from where the job was interrupted.

Note Image

You should run deduplication jobs when users are not modifying data on the cluster. If users are continually modifying files on the cluster, the amount of space saved by deduplication is minimal because the deduplicated blocks are constantly removed from the shadow store.

How frequently you should run a deduplication job on your Isilon cluster varies, depending on the size of your data set, the rate of changes, and opportunity. For most clusters, we recommend that you start a deduplication job every 7-10 days. You can start a deduplication job manually or schedule a recurring job at specified intervals. By default, the deduplication job is configured to run at a low priority. However, you can specify job controls, such as priority and impact, on deduplication jobs that run manually or by schedule.

The permissions required to modify deduplication settings are not the same as those needed to run a deduplication job. Although a user must have the maintenance job permission to run a deduplication job, the user must have the deduplication permission to modify deduplication settings. By default, the root user and SystemAdmin user have the necessary permissions for all deduplication operations.