How to do Readahead(file prefetching) Optimization

[adrian.oprea]

Hi everybody,
I have a question to make !
What is the optimal Readahead(file prefetching) value to be setup in Vertica ?

Sequential prefetching is a well established technique for improving I/O performance.

We know that HP Vertica recommends that we should set it to at least 2048
Example:

Code: Select all

/sbin/blockdev --setra 2048 /dev/sda1

Would my workload be good baseline for me to customize my Readahead value ?
Is there a best practice regarding this ?
thanx all

I found a very nice article on this topic and i would like to share it with you guys
http://www.ece.eng.wayne.edu/~sjiang/Ts ... dahead.pdf