Showing results for 
Search instead for 
Did you mean: 
   VIP    Certified

At the recently concluded Vision 2014, I had the opportunity to attend a hands-on lab session, Optimising Backup Exec 2014.  This lab highlights 4 tunable areas which can have an impact on the performance of your jobs.  We may come across these areas at different times, but this lab is the one which pulls everything together.  These areas are

  1. VMware backups
  2. Use of a dedicated network for backups
  3. Tape drive
  4. Dedup rehydration

The hand-out for the lab session is attached.  Although the lab uses BE 2014, these parameters are equally applicable for earlier versions of BE.  However, for these earlier versions, you got to check whether these parameters exists for your version.  For example, the VM parameter comes with Hotfix 176937 for Backup Exec 2010 R3, so if you are on BE 2010 R2 or BE 2010 R3 without this hotfix, then they would not be available.

Below are some additional references for these areas


You may be wondering.  With so many parameters, where do I start?  Do I change everything all at once?  How much improvement can I get?

There is one thing that you must always bear in mind when it comes to tuning, PATIENCE. When you tune anything, it is a slow and labourious process.  If you expect instanteous results, then forget about tuning your system.  You may get lucky, change a parameter and get 10% improvement, but how do you know you cannot get a 20% improvement.  Conversely, you change a parameter and your jobs slowed by 10%, but another change might bring you 20% improvement instead.  I am using the term, "system", instead of BE because the tuning methodology below is equally applicable to other applications or OS'es.

Establish a baseline

Before you tune your system,  you got to know the performance of the system at the moment.  You got to establish a baseline for your tuning exercise.  Your system has to be in a steady state before you embark on a tuning exercise.  If your system is not in a steady state, you would not be able to tell whether the improvement is due to your change or to a natural fluctuation in your system.

Once you have recorded your steady state, you can start tuning.  

Study your parameters

You need to do some research and understand what each parameter represents and do.  If you are tuning your tape drive, you got to understand what is thing called a block size.  What does it do and how would changing it affects your system.  Is there some dependency between it and the buffer size?  Some parameters are independent of each other.

Change ONE parameter

When you are tuning a system, you can only change ONE parameter at a time.  You cannot change more than one parameter at a time.  Otherwise, you would not know which parameter affect your system.  Also, you need to take baby steps, change one increment at one time.  Let the system settle down to a steady state and then record the result.  This may take some time.  If you are running weekly full backups, then you may need a few weeks before you can tell whether the system is in a steady state, i.e. the weekly backups are giving you consistent results.  This is why I said earlier that you must have patience.

The mathematically inclined among you would have realised by now that as the number of parameters increases, the number of combinations would increase exponentially.  This is to be expected.  Take tape drive tuning as an example. You would increase the block size to the next value and then change the value of the buffer size through its entire range to see the effects.  You would then increase the block size to the next value and run through the entire range of buffer size values.  This process is repeated until you have run throught the entire range of values for the block size.  My tape drive tuning exercise takes about 6 months.

If the parameter that you are tuning is not the limiting factor, then there will be little or no improvements when you change its value.

Choose the optimal set of parameters

From your results, you would be able to identify the optimal set of parameters for your system.  This optimal set of parameters is unique to your system at this moment in time.  For example, the values of a parameter may range from 1 to 9.  Your optimal setting for this parameter may be 6 whereas another installation may have an optimal setting of 3.  The other installation may get a 30% performance gain with its setting of 3 whereas you only get 10% gain with your setting of 6.  More does not always means better.  This is because there are other bottlenecks.  Going back to the tape drive, why can't you keep increasing the buffer size?  The bigger the buffer size, the more RAM you would use.  Very soon, you would be short of RAM.  Also, the tape drive may not be able to write out all the buffer in a reasonable amount of time, thus the performance does not improve.

Repeat the tuning exercise

Your system is not static.  Workload changes, you may put in more RAM or a faster disk drive.  Using the tape drive example earlier, if you are constrained by RAM and you just put in more RAM, then increasing the buffer size may give you some performance gain.  If you do not change the buffer size, then you would not get any improvement gain because the limiting factor is the buffer size.

Unless the performance of your system is very bad and you need some boost, tuning is most often done as a side-line to normal activities.  You tweak one area and then you tweak another area, and then come back and relook at the area again.  Tuning is a continuous exercise.


A good administrator will always find something to tune.  Happy tuning!


*EDIT - 27/9/2014*

You might want to use this GUI to help you tune your VMware buffers.




Level 2
Partner Accredited

Good Job !

Level 6
Partner    VIP    Accredited

Nice article pkh, good job!


Level 3
Employee Accredited

Hi Kok

Thanks for the superb article and explanations. Well done! yes

Toally agree that when it comes to tuning, PATIENCE is the key!

Level 0


Version history
Last update:
‎06-02-2014 07:09 AM
Updated by: