Video Screencast Help

Understanding VMWare Backups

Created: 24 Jan 2013 • Updated: 25 Jan 2013 | 3 comments's picture
This issue has been solved. See solution.

A couple of weeks ago I moved a batch of clients from typical client based backups to VMWare backups.  The elapsed time for each client is significantly better in most cases, but I have a couple of questions.  Here's the output for a 3 clients...

Client                     Duration                 MBytes                   # of Files

A - Before            1:13:46                     35,140                   309,816

A - After                 27:52                      56,010                  540,143


B - Before               32:52                     19,088                   30,139

B - After                 10:44                      29,366                  125,616


C - Before          01:41:41                      93,911                  230,408 

C - After             01:30:16                    116,684               1,807,396

My question is, why is there such a large discrepancy in the size of the backup, and even more so, in the # of files?  I love the faster backups but I wonder if I'm doing something wrong.

Thanks in advance for any insight you can provide.

Comments 3 CommentsJump to latest comment

William Jansen van Nieuwenhuizen's picture


Vmware backups is a very exciting backup feature within Netbackup and you'll see them ofthen punting the technology because it is really good. There are features you can enable, so that vwbackups become aware of changes on block level(something that's only available on normal clients if you are using nbu accelerator).

The last job, job c, seems a little wacky(size wise and speed wise). Are you using hotadd or San. Both good technologies if you set them up right. Hotadd can happen if you media server is also a vmware server(should say in the job details screen).

If this is the case, vmware should add the disk of other backup server via hotadd to your vm media server, eliminating the overhead of doing backups of the client over the lan first, before sending it to the media.

There is a lot of documentation out there, including the vmware netbackup guide. A quick google should provide you with enough features and sales material to keep you busy over a long weekend :-).

Yasuhisa Ishikawa's picture

It is from difference of level the agent take data on.

In stadard file backup, the agent on the machine read each file on the machine, and tranfer the data to  a media server as a sequence of files. Size of backup image is almost same with sum of backuped objects.

on the other hand, in VMware backup, the agent read VMDK and tranfer it as it is if no optimization is configured. If optimization is enabled, the agent pnly transfer *blocks* allocated for files on VM. even if some block is allocated for a very small file, whole data in this block will be backed up. So image size may become almost same with standard ALL_LOCAL_DRIVES backup in some case, but may not in some case.

Authorized Symantec Consultant(ASC) Data Protection in Tokyo, Japan

SOLUTION's picture


Thank you for the response, that makes sense.