Rank items based on different parameters Let's say I have N items to be ranked based on e.g. 5(it could be few or less) different performance measurements.
And ranking must reflect combined effect of all perfornance parameters.
Which stastical model should I use ?
I know this is too little information. But, I don't know what else I should write.
I can edit question as and when more clarfication asked by you guys.
EDIT:
Let me elobrate more and get into specifics.
I want to rank  my VoIP carriers in my application to optimize my call routing based on two performance parameters.
1) ACD(Avg Call Duration) = Total Duration / No of Successful calls
2) ASR(Avg Seziure Ratio) = No of Successful calls * 100 / Total Calls (Including failed calls)

I measure above parameters from time to time from accumulated past Call Detail Records.
Usually the sample period is 6 hours.
I may consider ranking individual dimensions and then averaging them. But,I don't want a particular dimension to dominate. That means, if a carrier provide good ACD but bad ASR - it's bad for me. Vice versa.
 A: One approach is to compute the z-score for each performance measurement. Calculate a weighted sum of the z-scores to arrive at a composite z-score and use this to rank order your observations. This composite z-score not only provides you with an ordinal way of ranking the observation but it also provides you a way of assessing each observation relative a standardized scale.
The important part here is assigning a weight for the calculated z-score of each performance measurement. This should solve your problem about a particular metric dominating the others.
A: I could be wrong, but I think you need to decide the relative importance of performance parameters based on domain knowledge. At least in my knowledge statistics offers you no insight into the relative weights. 
e.g. Say you had to rank cars on the basis of gas-mileage and top-speed. Every persons mental model of car rating would give different weights to each of these two parameters. So I am doubtful you can get Statistics to solve this conundrum for you.
I agree with @BGreene that PCA might help you understand the dataset; though high variance by itself does not imply that a variable must be the most important in a rating context. 
Again, I could be wrong. 
