du performance

[Mag0007]

if i have a very large filesystem (/usr) is there a way to improve the performance of du /usr?

it takes about 5 mins to get my results back...

 

[Annihilannic]

I don't think you can make it much faster. Alternatively you could use find /usr -ls and add up the file sizes using awk or something, but I'm not sure whether that will achieve what you require.

Annihilannic.

 

[Mag0007]

thanks for the reply. Yes, you are correct I don't think it will really acheive what I require, because I will be doing a lot of pipes with du (pipes to sort and awk)

 

[stefanwagner]

But a second invokation is much faster.
There must be some clever caching.

seeking a job as java-programmer in Berlin:

http://home.arcor.de/hirnstrom/bewerbung

 

[Mag0007]

stefanwagner:

VERY VERY true! on AIX it works like that.

Thanks for the reply!

 

[p5wizard]

You can improve the perf. of your script even more by first storing the output of du in a temp file and instead of doing du over and over just cat-ing the temp file on the sort and awk pipes.

It's a tradeoff between accuracy and strain on IO-system, because you'll be basing your sort and awk pipes on the du output of a while ago, but catting a file is easier than stat-ing each file/dir in a filesystem and counting blocks - even if the data is 95% cached...


HTH,

p5wizard