|
Home > Archive > Data Storage > July 2005 > How to predict the I/O workload on demanding LUN
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
How to predict the I/O workload on demanding LUN
|
|
|
| Dear friends,
Is there any way to dynamically predict the I/O work load for
different mode means which LUN is utilising maximum.I mean to say
suppose many applications are running on the host.But based on the time
I want to know the I/O worklaod which has the maximum. I have to take
consideration like logging,file system logging which are less important
for me. I donot have to predict it manually but I specifically want to
know dynamically how the load is on the different LUN from the host
perspective.
Thanks & Regds,
BABI
| |
| Nik Simpson 2005-07-05, 8:45 pm |
|
"babi" <tmohanta@gmail.com> wrote in message
news:1120559773.743621.180350@f14g2000cwb.googlegroups.com...
> Dear friends,
> Is there any way to dynamically predict the I/O work load for
> different mode means which LUN is utilising maximum.I mean to say
> suppose many applications are running on the host.But based on the time
> I want to know the I/O worklaod which has the maximum. I have to take
> consideration like logging,file system logging which are less important
> for me. I donot have to predict it manually but I specifically want to
> know dynamically how the load is on the different LUN from the host
> perspective.
In order to make any sort of meaningful prediction about future performance
you need to gather IO stats for the system under typical load. Since (based
on your description of your needs) you want to relate the IO with the
processes running on the host you'll probably need to gather stats on the
host not the array (assuming the array even allows gathering of performance
information.) The procedure for capturing IO stats on the host is very much
OS dependent and since you didn't say what OS you are using it's impossible
to give any further advice.
PS. It might help everyone if you could start with a high level description
of what you want to achieve and why. Without understanding your goals its
only possible to give answers to the questions you ask instead of perhaps
suggesting alternative approaches that would get the job done.
--
Nik Simpson
| |
|
| Hi Nik Simpson ,
Thanks for your input and suggestion.
My requirement is simple like as follows.
In load balce we have round robin, shortest path etc are there.
Similar I want to have a priority based LUN mean the LUn which have the
highest I/O or work load will having the first priority in the quqe to
send to the storage device from the host. As a result it will get more
attention as compared to others. Look at the scenario as L1,L2,L3,L5,L6
are 6 luns. At any point any one will have highest work load and that
point the processing should be changed and that particular LUn will
serve the purpose. We want to implement this on Linux OS first.
(1) Dynamically we want to know the work load or I/O for the lun
considering other aspect also like logging of the file system etc.
(2) manually if I know the load behaviour I can set for that time with
peak I/O to acheive or that lun will have higher priority etc.
But we are keen to ponder on the idea (1) to acheive dynamically.
Your suggestion and help is appriciated.
Hope this is clear and specific.
Thanks & Regds,
BABI
| |
|
| I=B4m curious about which array do you have. Some vendors claim that
they have automatic load balancing and performance managemente made by
algorithms.
regards,
JP
| |
| Nik Simpson 2005-07-13, 5:52 pm |
|
"babi" <tmohanta@gmail.com> wrote in message
news:1120802900.081431.189460@o13g2000cwo.googlegroups.com...
> Hi Nik Simpson ,
> Thanks for your input and suggestion.
I was looking for a description of the application that is generating the
I/O workload that is leading to the questions you are asking. I can't speak
for anybody else, but I'm afraid I can't understand your more low level
problem descriptions due to language difficulties. Maybe if you could start
from the very highest level we might be able to drill down, but starting
from the bottom up isn't working for me.
--
Nik Simpson
|
|
|
|
|