NetBackup 10.5 Offers High-Speed Multi-Streamed Snapshot/Replica NAS Recovery
In NetBackup 10.5 we’re introducing parallel recovery from NAS snapshots and replication copy backups. In previous releases we offered parallel recoveries from NAS backup images. Now we have this feature for NAS snapshot and replication images as well. To use the parallel recovery from snapshot feature, users need to set the snapshot or replication copy to be restored as the primary copy. They then run a restore job with a “number of streams” value set to be greater than “1”. An example number of “7” is shown in the figure below. A matching number of streams are produced and sent to the filer. These streams are spread as evenly as possible across the pre-configured backup (mount) hosts available (backup host pools or individual backup hosts). If only one backup host exists, all streams will come from that host and may cause disk and network I/O resource contentions if the number entered is too large. Multiple backup hosts offer better throughout and eliminate single points of failure. You’ll want to test your restore configuration accordingly. If you need to add backup hosts, see the backup host creation/configuration information in the NetBackup Snapshot Client Administrator’s Guide. NetBackup 10.5 also adds Checkpoint Restart (CPR) to NAS restores. CPR is supported for both single- and multi-stream restores. This means NetBackup administrators can now manually suspend and resume their restore jobs when necessary. (This is a critically important feature on production filers performing work and being restored at the same time.) Failed jobs can also be resumed after the failure causes are eliminated. CPR is supported for the backup copy only. It is not supported for restoring from snapshots or replica copies. NetBackup automatically maintains check point intervals as restore jobs progress. Architecture The graphic below shows the NetBackup processes used in a parallel NAS/DNAS recovery job from snapshot or replica copy images. The primary server coordinates the job between the mount hosts reading the image and destination filer host TAR-writing the recovery data. Each mount host mounts the snapshot(s) required and then streams the data to the NAS filer. TAR on the filer then writes the recovery data to the filer’s storage. Each stream is shown in Activity Monitor as a separate job under the parent restore job. Requirements and Prerequisites The following are needed to use NAS parallel recovery streams and CPR restores: Any supported NetBackup 10.5 primary server, media server, and NAS filer(s) Each restore host operating system must be RHEL, SUSE, or Windows A NetBackup 10.5 Snapshot Manager server-supported operating system with the following or newer: - Ubuntu 16.04 and 18.04 Server LTS - RHEL 8.2 and 7.x 10.5 Performance Increases As shown in the table below, parallel multi-stream NAS recoveries produce a significant restore performance improvement over single stream restores. The following example test was performed in our labs on a volume having close to 600K files of almost 26GB. Backing up the volume using DNAS with five streams took about eight minutes. A single-stream NAS restore required 27 minutes to finish the restore. A multi-stream NAS restore with five streams took close to nine minutes. That’s barely one minute more than the original backup, and almost unheard of with previous NAS restores. This means this new feature makes the restore time almost the same as the backup time using the same computing resources. Conclusion This great new NetBackup NAS multi-stream snap/replica restore feature can really save DR time and headaches. Administrators no longer have to dread filer restores if they use this feature. And all that’s required is using the right primary copy and specifying the number of mount hosts you have available for the job. Upgrade to NetBackup 10.5 today and lose that filer restore anxiety.
