Appliance backup lower than 20MB/s

[sistemas_ctc]

In customer we have:

1 switch Procurve 1Gb/s

1 Appliance 5230 like media server with RAID6 (14TB) MSDP. 2.6.0.3 Version  (name HOLLY)

1 Master Server with Netbackup 7.6 (name BEN)

All licensed required... to acelerator, NDMP, deduplication, MSDP, OpenStorage..etc

File System is a CIFS shared  from HNAS404 with name italica

----------

We have mount italica in holly by CIFS and NFS:

holly:/home/maintenance # cat /etc/fstab
10.231.192.131:/sscc            /mnt/italicaNFS                 nfs     defaults        1 1
\\10.231.192.131\sscc           /mnt/italicaCIFS                cifs    credentials=/root/.usuarioCIFS 1 1
holly:/home/maintenance #

policy have Holly like client, type standar, acelerator enabled, and /mnt/italicaCIFS/xxx like backup selection

----------------

with symantec the job finished at 8 minuts, in the switch we detect only 25% of traffic bettwen HNAS and holly

we probe to copy in holly (cp -r /mnt/italicaCIFS/xxx /home/prueba/) the same folder and the copy finished at 2 minuts and switch detect traffic 100% in the ETH of holly and 70% in both eth of HNAS

----------------

someone can help with this??

we modify the buffer configuration, launch first FULL backup with "force rescan", we try to mark "multi stream" and the windows of backup it's the same.

xxx its only one folder of the filesystem, a FULL backup of italica finished at 40h!!! with force rescan (2,8TB and 2.760.000 concurrent files)

[chashock · Accepted solution]

To be clear on this, you know that mounting a CIFS share to the appliance is not supported, right?  You're trying to do something that hasn't been tested or approved as a use case on the platform, and it certainly hasn't been performance tested by Engineering.

Is this a multi-protocol file system so that you could try mounting it via NFS (given what you've noted I'm assuming it is).  What results do you have backing it up over NFS (which is supported)?

A backup job isn't the same as a filesystem copy, so I don't expect to see the same results on network utilization or speed.  There's more going on here with a backup.  Do you have client dedupe turned on?  

For accelerator, you don't want to check the box on your normal full schedule to force a rescan.  Your first full will have that happen regardless.  Just create a second full schedule with that check box enabled with no dates to run it for accelerator to function properly.

Do your job details indicate any acceleration is happening?

You have a lot of files in 2.8TB, CIFS is a terribly slow protocol, and in my experience HNAS isn't the fastest NAS platform out there, so depending on how it is configured you could have issues on the storage side as well.

There are a lot of variables here to consider for making your backups faster.

[watsons]

Just curious, why do you choose to backup via CIFS (mount on a client), rather than using NDMP policy type since you have that license too?

HNAS is supported as NDMP device in compatibility guide.

I have tried doing something similar in Netapp CIFS mount, it's terribly slow and prone to network connection failure. So I don't really like backing up CIFS share.

[dbeckford]

I am interested to know what ever happened with your issue? Were you able to get better performance?