Further Information

Performance Considerations

The speed also varies if a device-compression gets used. The achievable compression ratio depends on the nature of the data being backed up. For most cases, using high speed devices with device-compression ON does improve performance. This however is true only if the device(s) stream.

Libraries offer additional advantages because of their fast and automated access to a large number of media. At a backup time loading new or reusable media is needed and at a restore time the media which contain the data to be restored need to be accessed quickly.

High Performance Hardware Other Than Devices

The computer systems themselves, that is, reading the disk and writing to the device, directly impact performance. The systems are loaded during backup by reading the disk or handling software (de-)compression.

The disk read data rate and available CPU are important performance criteria for the systems themselves in addition to the I/O performance and network types.

Using Hardware in Parallel

Using several datapaths in parallel is a fundamental and efficient method to improve performance. This includes the network infrastructure. Parallelism helps in the following situations:

•Several systems can be backed up locally, that is, with the disk(s) and the related devices connected on the same system.

•Several systems can be backed up over the network. Here the network traffic routing needs to be such that the datapaths do not overlap, otherwise the performance will be reduced.

•Several objects (disks) can be backed up to one or several (tape) devices.

•Several dedicated network links between certain systems can be used. For example, system_A has 6 objects (disks) to be backed up, and system_B has 3 fast tape devices. Putting 3 network links dedicated to backup between system_A and system_B is a solution.

Appendix A

A-9