Poisson?

[Martin Payne]

Hi all

I'm hoping that you can help me with a statistical problem. I am load testing a system and have found that, when X number of transactions are performed concurrently, performance degrades (no need for specifics on the degradation). What I need to do is identify this in my report and then indicate the probability that X number of transactions will be performed concurrently, expressed as "once every Y hours" (it's a report for management so I need to keep it simple)

My inputs are that I know how many transactions will be performed every hour (say 'n') and the length of time it takes to execute a transaction (say 't'). Now, I want to keep it simple so I'm not too concerned about transactions that start at slightly different times yet, due to t, overlap. In this case I can just assume that they are running concurrently and started at the same time (so, for a transaction that has t = 5 mins, there are only 12 possible intervals in which it could run)

I've had a look at the Poisson function but, firstly, I'm not sure that it applies to my requirement and, secondly, I'm not sure how to input my figures. Any advice

Thank
Martin

 

[Phan]

If there are 12 possible intervals, the average in any one
interval is n/12.

Just use the Poisson function as

=Poisson(X, n/12, False)

-----Original Message-----
Hi all,

I'm hoping that you can help me with a statistical

problem. I am load testing a system and have found that,
when X number of transactions are performed concurrently,
performance degrades (no need for specifics on the
degradation). What I need to do is identify this in my
report and then indicate the probability that X number of
transactions will be performed concurrently, expressed
as "once every Y hours" (it's a report for management so I
need to keep it simple).

My inputs are that I know how many transactions will be

performed every hour (say 'n') and the length of time it
takes to execute a transaction (say 't'). Now, I want to
keep it simple so I'm not too concerned about transactions
that start at slightly different times yet, due to t,
overlap. In this case I can just assume that they are
running concurrently and started at the same time (so, for
a transaction that has t = 5 mins, there are only 12
possible intervals in which it could run).

I've had a look at the Poisson function but, firstly, I'm

not sure that it applies to my requirement and, secondly,
I'm not sure how to input my figures. Any advice?