Executive Summary
This white paper provides
This white paper covers HP ProLiant Windows 2003 server connected to HP StorageWorks Modular Smart Array (MSA) disk arrays and different tape drive technologies (LTO and SDLT).
The proof points are all
As a result of these tests, several recommendations and rules of thumb have emerged:
•HP OpenView Storage Data Protector tuning can help to improve the performance, e.g. by modifying tape block and file depot sizes. Please check chapter Tuning Recommendations.
•HP StorageWorks Ultrium 960 tape drives are best utilized with a block size of 256 KB as described in “Getting the most performance from your HP StorageWorks Ultrium 960 tape drive white paper” (downloadable from http://h18006.www1.hp.com/storage/tapewhitepapers.html).
•A single
•For
•The configuration of disk arrays can have a remarkable impact on the backup and restore performance. Important parameters are the configured number of logical arrays, logical volumes and type of RAID levels. The configuration of logical arrays and volumes should reflect the internal disk array layout. E.g., disk arrays with 2 SCSI buses could perform best by creating 2 logical arrays with one logical volume each. RAID levels must be chosen carefully. RAID 0 provides the best performance but should not be considered due to missing fault tolerance. Therefore, it is recommended to configure RAID 5, which fits best for staging areas. It is space and cost efficient and provides a good read performance.
•Serial ATA (SATA) disks used as secondary disk storage arrays can have slower rotational speeds than their SCSI counterparts (for example, the HP StorageWorks Modular Smart Array SATA drives are only 7.2K rpm) and although they are high capacity (250 GB), their performance reflects their pricing. Therefore, backing up to tape from a staged (secondary) disk array can be slower than backing up directly from the primary storage to tape in some circumstances. Ironically, tape is now faster than disk. Disk staging is useful, however, for gathering several small files into a single object, or backing up slow networked hosts before the data is sent to tape. Both small files and slow hosts can cause very slow backups.
•The restore of many small files (22 million in this setup) could cause serious file system bottlenecks. Data Protector must wait until Windows and the belonging NTFS responds.
•Finally, the test environment with the HP ProLiant Server DL380 G4 and MSA1000/1500 disk arrays is capable to manage data with low CPU utilization. The typical file backup (from MSA100) directly to the Ultrium 960 tape drive showed a high transfer rate (91,97 MB/s or 323,33 GB/h) together with a low average CPU load (11%). The comparable disk backup (from MSA1500) to tape still showed an acceptable transfer rate (43,37 MB/s or 152,42 GB/h) together with a low average CPU load (9%).
3