HP StorageWorks 6000 HP StorageWorks VLS and D2D Solutions Guide (AG306-96028, - Page 160

Media Management To make the most of storage savings with Accelerated deduplication, consider the following: • Reclamation does not occur until a tape is full (and everything on the tape is either delta-differenced, disabled, or unsupported). • The deduplication software cannot deduplicate versions of a backup that are on the same cartridge; the versions are not deduplicated until a new version is written to a different virtual cartridge. Use the following guidelines: • Appending media pool HP recommends using standard, appending shared media pools across multiple tapes. By doing so, tapes fill with backup jobs from different hosts, which allows deduplication to occur and storage to be reclaimed. You should not use non-appending media pools. • Virtual cartridge sizing Generally it is always better to create smaller cartridges than larger ones because it frees up the cartridge quicker for use by other processes such as restore, tape copy or deduplication processing. Cartridge sizes of 100-200 GB are common and this size can be used regardless of the tape emulation type because the cartridge is virtual (i.e. you can create a 100 GB cartridge even if you selected LTO4 drive emulation). However, with Accelerated deduplication there are other sizing considerations that could recommend even smaller cartridge sizes. If the daily amount of backup data written to your cartridges is significantly less than the cartridge size, the deduplication process will only be able to eliminate the duplicate data (i.e., free up disk space) once the tape has been filled up by multiple days of backups. In other words, if your average backup job size is significantly smaller than the virtual cartridge size: • Divide the number of jobs by the number of virtual tape drives. This determines approximately how many jobs will write to each cartridge per night and thus how much logical data is written on average to each virtual cartridge. • Divide this logical data per cartridge by the average data compression ratio to get the native capacity required. • Compare the native capacity required to the cartridge size. If the native capacity written per night is less than the cartridge size, reduce the cartridge size to roughly match. For example, if you have 50 backup jobs per night with an average size of 25 GB each, writing to 10 virtual tape drives, this writes 125 GB logical data per drive (50/10 * 25). With 2:1 compression you get 62.5 GB native disk capacity per drive per night. In this case, use a cartridge size of 50 GB. The recommended virtual cartridge sizing for most environments is 50-100 GB. You may need to configure a subset of your cartridges at a larger size if some of your backups are very large and you want to reduce the number of tapes the backup spans. (Spanning too many tapes reduces deduplication performance.) The maximum recommended cartridge size is 300 GB. Client and Backup Job Naming Deduplication relies on the names of the backup job and the client to correlate them. • If you rename the backup job or client after you have deduplication running on your system, the deduplication for that backup data is reset (as though you just activated it for the first time). 160 Virtual Library Systems