There are many factors that can limit overall backup performance, potentially causing backups to run into production windows, reducing backup success, or increasing infrastructure costs.
Many backups are limited to some degree by LAN/SAN bandwidth available to the server being backed up. Adding the backup overhead inefficiency of the CIFS and NFS protocols over Ethernet further exacerbates the problem. Other backups are limited by the speed and number of available Fibre Channel connections, due the expense of having to upgrade to the latest port speeds every few years. Another common choke point for traditional backups is overworked backup servers, especially during full backup periods.
Administrators also frequently have to deal with a limited number of slow and unreliable tape devices that restrict backup performance. In addition, tape media failures reduce backup success and cause additional work for your staff rerunning failed backup jobs. Finally, with physical tape drives, if you can’t keep the data streaming fast enough, the tape buffer empties and the tape drive dramatically slows down. This reduces overall backup performance and causes premature wearing of the tape drive head—further reducing backup and restore reliability. Backups using physical tape of virtual tape also create the additional challenge of managing potentially thousands of tape cartridges associated with day-to-day backups and disaster recovery.
There is a lot of change occurring in the data protection landscape today. Even with purpose-built backup appliances (PBBAs), simply replacing tape is not enough. For application owners such as DBAs, difficulty meeting backup windows due to an explosive growth of data that needs to be protected, and lack of control and visibility into their own backups, has resulted in dissatisfaction. Application owners have other alternatives for protecting their data; when they don’t get the control and visibility they want over their own backups, they go off on their own, leaving the IT team with less visibility and control than ever before. As a result, this can lead to unnecessary silos of storage, or what can be called an accidental architecture.
EMC’s Data Domain deduplication storage system performs 99% of the deduplication effort in CPU and RAM. This means that Data Domain systems are CPU-centric, and not spindle-bound for performance like other deduplication platforms. This is also why Data Domain systems have achieved dramatic increases in overall performance with each successive generation using the latest Intel processors.
Visit siriuscom.com/goEMC to learn more about EMC solutions from Sirius, or contact us for more information.
