backint interface

[Former Member]

Hi team,

Can anyone throw some light about the backint interfaces for direct backups rather than disk based backups.

How is the performance when compared to the backups to disk and then take it forward from there.

Also how do we analyse the backup performance we need metrics from linux team if the backups are slow right ?

What would be a good starting point ?

Message was edited by: Tom Flanagan

[Former Member]

Hi,

We have got our backups setup to both local NFS share and Backint tool to tapes

Based on the infrastructure we have in place the backups are about 50% slower to the Backint Tape compared to the NFS share

However it is still a good option because the backups when taken to NFS share need to be copied to tape at a later time and Backint process is a one step job

Regarding your latest question for restore:

You would not need the Config files for a restore if the source and destination is the same system

You would need the config files to be checked and set to the same values if you are restoring from a Source to a different Destination system

Our Backint tool documents not to recover the backup from one source to another destination as it is not supported, so we use local backups for refresh from PROD to Non Prod(we already make sure the config files are identical across all systems before restore)

[Former Member]

Hi,

I have a customer using Netbackup and backint for data and log backups, the throughput is around 50 MB/s (untuned) , whereas a disk backup is significantly faster it does require a 2 stage backup (-> Disk -> Tape) and the disk storage to be available at all times. For disk based backups I have see from 100MB/s to 400MB/s depending on the storage type used (NFS vs. Local)

Hope this helps.

Thanks

[Former Member]

Paul,

Thanks a ton for the info.

So that concurs backing to the external backup solution is slow, backing to disk is faster and we can later on back it up from the FS.

How does the logic work with the backup solutions , does it put a write suspend for the backup to make system consistent since its a multinode or else since all the index servers across all the hosts should have the savepoint number to maintain consistency.

Thanks,

[Former Member]

Hi,

I wouldn't say it is slow since you can spend the money on network and tape infrastructure to get the desired throughput, modern backup technologies can back up to disk that is controlled by the backup software and then move it to tape at a later stage (yesterdays on disk, older than that on tape)

, disk is almost always going to be faster using current technologies. Some advantages of using a 3rd party backup solution is that they backup the config files and the catalog files required for a recovery , use compression on the backup files since they are uncompressed on disk, avoid disk full issues

For the detailed logic I would refer to official documentation (below) but basically the backint initiates the backup on all nodes and puts the database into backup mode (creating its own savepoint), backs up all the files it needs to and then completes the backup.

For log backups, once a log buffer is full or committed it is written to the log segment in the persistent layer, once the log segment is full or the time threshold is reached it is backed up to tape and released (only after the statisticsserver process has confirmation of the backup)

For my customer we used Netbackup, so I am referring to that documentation.

Admin Guide - http://www.symantec.com/docs/DOC6482

White Paper - http://www.symantec.com/docs/TECH208919

Thanks

Paul

[Former Member]

Paul,

One question before i respond to your points.

By how much factor did you observe a reduction in the space usage when compared to disk usage by the backup.

Thanks,

[Former Member]

Paul,

By what factor did you obeserve a reduction in the backup image size.

suppose we have a 1tb backup size on the disk , what happens when it goes directly to netbackup? What would be an approximate size?

[Former Member]

Paul,

Also did you observe backup being slow when it was written to NFS disk.

Did you have multi node or single node?

Which appliance were you using ?

[Former Member]

Paul,

One question when restoring do we really need config files which had been backed up?