Tuning NetBackup: Benchmarking Clients

Question:  How fast can I backup? Answer:  As fast as you backup. I often joke about this common NetBackup support question.  Though this is my candid and brutally honest answer, there is a way to benchmark what you could expect from backing up a particular client by examining the performance of the bpbkar process.  Though there are many factors that impact the overall backup performance of a specific client this is a good starting point. For this example I will show you how to benchmark the bpbkar32 process on Windows clients. First you will want to make sure that your windows client has the NetBackup agent installed and the bpbkar logging directory created à  we will need this to analyze the results.  Example logging directory = C:\Program Files\veritas\netbackup\logs\bpbkar You will then want to run the following command (path may be different if NetBackup is installed on a different drive on the client): C:\Program Files\veritas\netbackup\bin\bpbkar32.exe –nocont E:\ 1> nul 2> nul (bpbkar32 –nocont [file_path_to_test] 1> nul 2> nul) Once the command is executed make sure it runs to completion by monitoring processes and making sure the bpbkar32 process is not running or refreshing the bpbkar log file until it does not increase in size.  After full execution open the log located in C:\Program Files\veritas\netbackup\logs\bpbkar and find the “elapsed time” entry and record the bps listed at the end of that line.  In this fictional scenario I recorded 31258568 bps. Quickly converting the 31258568 bps (divide by 1024 for Kilobytes and again by 1024 for Megabytes) you come out with 29 Megabytes per second. Question:  Well what do I do with that? Answer:  We‘re not there yet. The reason that “We’re not there yet” is that we must do this a few more times to gather additional benchmark snapshots.  Additionally you should run this on large datasets that are or are similar to the datasets that we will be backing up on a regular basis. Question:  Okay, so now I have run this many times on similar large datasets, now what? Answer:  Once you have this data you know “theoretically” how fast you will backup that particular client. Question:  “Theoretically”, well why did I just do all of this? Answer:  Okay so “theoretically” is a strong word, I just want to set the appropriate expectations as in real world scenarios there are other factors that impact backup performance from network interface cards, cabling, switches, media server performance, tape performance, virus scanning, third-party applications, number of files,  if Mars is aligned with Venus.