What are the steps to optimize the file server backup

Before you do optimizing, you better test/verify your environment first.

I assume you have millions of smaller files on your system

If your tape drive runs the maximum transfer speed, this is it.

If not, it could be anything else. We do not know your congiguration.

Have a look at the Performance Optimization Planning Guide - it will give you a good start where to look at.

Thanks for your reply Bingo, you are right we have millions of files in the backup directory since its a file server I have already done the necessary optimization according to the document. Our tape drive through put is 300 MB/s when configured it as storage node for other application backups(SAP& LOTUS). so its not related to speed of tape the bottle neck of the backup is it consists of millions of files and folders. In earlier networker version a module called SnapImage module later is turned to NSM which requires storage Array or DATA DOMAIN Snap image module creates a virtual NDMP on the client so that the client backup is work like NDMP Dump backup refer the link below and give me a suggestion

Link :  Tech Tip: Backing up High Density File Systems

   "Our tape drive through put is 300 MB/s when configured it as storage node for other application backups(SAP& LOTUS)."

Why do you state that? - That's nice for DBs but how does it actually perfom with such file monsters?

Think about 'the tape drive’s view':

  To work optimally during backups, it just needs data to keep on streaming.

  It cannot go faster than the raw data rate allows.

  However, if the data rate is lower (most likely for a lot of small files), there are 2 options:

     - it can produce a longer gap and keep on streaming, loosing media capacity or

     - it can stop and reposition the tape, which takes time (some seconds) and which will decrease the throughput speed even more.

Actually, I don't know whether the first method does still work with tape drives today (in fact, it was never often used ;-) .

So what can you? - In general you have to test where to improve:

  - On the source:

      Invest in faster disk systems

        (for example in our environment the backup times did reduce by about 60% just when we replaced our VMAX2 with a

        VMAX3, backing up to a DD4500).

  - On the transport media:

      - A 10gBE/SAN should be fast enough

      - Maybe you can bundle some channels

  - On the backup media:

     - Buying more tape drives ? - does not make sense if you already can't keep a single one streaming.

     - Backup-to-disk ? - most likely better to reduce latency

     - Backup-to-DDD ? - the best method to benefit if you also use DDBoost

  - On the method:

     - Zipping before backup - but this needs time and resources aas well

     - PSS - needs more backup devices

     - BBB - I am not shure thether it can be used along with tape drives (it does not make sense, of course)

     - NDMP backups - if you can move the files to an NDMP client

         - you may also think wrapping NDMP data using DSA (Data Service Agent)

     - SnapImage - I think it is still around but in general it is nothing else as 'a kind of NDMP backup for Windows filers'

Of course, combinations may also be possible

The conclusion:

  - First, change to B2D (see above)

      - clone from here if you need the data on tape

  - Think about the backup method

  - Make proper tests and argue with your management to get the best solution for your scenario.

      Don't think about the price - it will pay in the future.

Good luck

BTW ... I forgot that your tape drive just cannot stream. Why? - simply do the math:

  300 MB/s * 3.600 s = 1.080.000 MB/h

Consequently, your tape should finish the 3TB backup job in roughly 3 hrs. But it takes 30, you mentioned.

But heads up: we have one linux client with 3 major file systems. In total

  - 4.4 TB

  - 18.3 Mio Files

Backup time to DDBoost: about 11 hrs - and this is a standard filesystem backup

And this is only one machine among 600 clients.