Message-ID: <29349960.1075856827441.JavaMail.evans@thyme>
Date: Fri, 19 May 2000 08:35:00 -0700 (PDT)
From: chonawee.supatgiat@enron.com
To: vince.kaminski@enron.com
Subject: XPRESS and CPLEX Runtime Comparison results
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-From: Chonawee Supatgiat
X-To: Vince J Kaminski
X-cc: 
X-bcc: 
X-Folder: \Vincent_Kaminski_Jun2001_7\Notes Folders\Discussion threads
X-Origin: Kaminski-V
X-FileName: vkamins.nsf

Vince, 
FYI
below is the comparison of the computation time results from CPLEX and XPRESS 
optimization softwares.
-Chonawee


---------------------- Forwarded by Chonawee Supatgiat/Corp/Enron on 
05/19/2000 03:26 PM ---------------------------


Chonawee Supatgiat
05/17/2000 05:48 PM
To: samer_takriti@enron.net, Stinson Gibner/HOU/ECT@ECT, Ravi 
Thuraisingham/Enron Communications@Enron Communications
cc: Tom Halliburton/Corp/Enron@Enron, Pinnamaneni Krishnarao/HOU/ECT@ECT, 
ming-lung.lee@sycamorenet.com, ming888lee@aol.com 

Subject: XPRESS and CPLEX Runtime Comparison results

Hi,

I have tested both CPLEX and XPRESS-MP optimization software in solving our 
network design problem.
XPRESS is more user friendly hence it takes less developing time. However, 
CPLEX performs significantly better
in term of computation time, especially in integer programming.
As a result, eventhougth XPRESS is easier to use and cheaper, I would 
recommend our EBS group to use CPLEX. 
Please let me know what do you think...

The test problem has 10 nodes, 77 cables, 5 cable speed. 


The table below represents the computation time in seconds for solving this 
problem.
The second column indicates if the general integer variables are relaxed (but 
still keep the binary variables). 
The "LP" column represents the computational time when solving relaxed 
problem (no binary & integer variable constraints). 
XPRESS took too much time in solving case 4 and 6 so I had to stop the 
process and report the best solutions it found.




Below are the graph.



