Unsolved
This post is more than 5 years old
60 Posts
0
3729
Backup is too slow with large save set with more than 10 million files
Dear All,
I have a backup client consist large save set with more than 10 million files
Operating system : Windows 2008 R2
The Backup is too slow that its running in KB/sec speed even in SAN and LAN backup
There are no server utilization while backup is running , is there any way to increase the backup speed when its running for million files
Kleinenbroich
86 Posts
0
October 27th, 2015 06:00
Hi,
you can try to use the bigadm directive to check whether really the file system is the bottle neck:
https://emc--c.na5.visual.force.com/apex/KB_HowTo?id=kA0700000004KjG
Regards
Michael
bingo.1
2.4K Posts
0
October 27th, 2015 11:00
First of all it is important that you use B2D - tape drives are no good as they permanently shoe-shine with these data rates.
Then, you do not have much choices:
- either you compress (zip) the files at the client
- or you run an image-style backup (block based backup or RAW partition backup)
This is pure NetWorker.
It will speed up if you think about a combination of NW and Avamar but this of course will need extra investments.
crazyrov
4 Operator
4 Operator
•
1.3K Posts
1
October 28th, 2015 01:00
I would suggest you to make use of Block based backup which is designed to solve problems similar to what you are currently facing.
however this solution required the target device to be a disk type (AFTD or DataDomain), you can propose to add disk storage to your backup infra to create a AFTD and do a POC for Block based backup for your organization to approve the additional charges.
Vakil
35 Posts
0
October 28th, 2015 03:00
I also do agree of leveraging BBB. There could many causes also.
Bigasm can help knowing the bottleneck.
You can also divide into many client policies and run backup for individual savesets, and not all at once.
Kleinenbroich
86 Posts
0
October 30th, 2015 08:00
Hi Bingo,
please keep in mind that "compress (zip) the files at the client" makes only sense if neither ddboost nor deduplication is used (e. g. in case of having a DataDomain backup device).
Regards
Michael
bingo.1
2.4K Posts
0
October 30th, 2015 14:00
@vakil
Actually, bigasm is more testing network and backup device speed than anything else - so I think the results are not valid in this case.
@kleinenbroich
I am not sure about ddboost - we use it on similar filesystem with no client compression and the throughput is rather slow as well. But keep in mind that it all depends on a lot of factors, so you must run true tests with your true data and your backup scenario anyway.to find the best method.