Netbackup 7 Launching : De-dupe and Virtualization Backup

Symantec launched its Version 7.0 Netbackup at Nikko Hotel today.

Netbackup 7.0 key points:



1. De-dupe
Netbackup 7.0 focus on the idea of de-duplication everywhere. Using de-dupe to significantly cut down the volume to be actually backup/stored. At the same time reduce the resource utilized, e.g. reduced Network traffic, reduced WAN traffic for DR site replication, reduced storage space, reduced backup time.

Of course with de-dupe introduced, some CPU and RAM overhead needs to be accounted for, and it depends on where you decided to actually run the de-dupe algorithm.

a. De-dupe appliances - Data sent from client to media server then onto a de-dupe device to actually do the de-dupe algorithm. Resource is only saved at the final stage when "Unique Data" is written onto storage.

b. Media Server - Data sent from Client to media server and then do a de-dupe at the Media server. Then "Unique data" is written onto storage. Savings is between the media server and the final storage.

c. Client side - Data is de-dupe at client side before heading out to the Media server. Network utilization reduced, Storage required for backup is reduced, backup window is reduced, but some overhead at the CPU and RAM over at client side needs to be sized in.



2. Virtualization Backup.
Apparently Netbackup 7.0 is able to do virtualization backup more efficiently than other competitors. Reason being Netbackup 7.0 only needs to backup once on the vmdk to be able to do both "Whole VM" restoration and "Single File" restoration, whereas other vendors needs to do 2 backups to accomplished that.

A new term "Change Block Tracking" is also mentioned during the briefing. Went to this blog and got some idea about CBT. (Seriously good info for Virtualization and Cloud)

CBT seems to be a new features for Vsphere 4.0 kernel. Instead of writing own code to pull block info from the vsphere kernel, CBT allows the 3rd party apps to query the info directly from the kernel and the result is better efficiency.

It is also used in VSphere 4.0 VMotion where it works like a buffer area to keep track of all the I/O while Vmotion is copying the initial copy of the data files. After initial copy of data files are complete, the CBT will be traced to write the changes into the initial copy. Where at the same time new I/O are traced continuously.

The algorithm goes on until the changes are really small and the copy can be done instantly. The VM will then be suspend/resume (instantly) after the final traces are done. This I guess gives even seamless Vmotion.