Take the 2-minute tour ×
Geographic Information Systems Stack Exchange is a question and answer site for cartographers, geographers and GIS professionals. It's 100% free, no registration required.

First of all this is my first question on this site :) And I am really happy to be within the community.

I have no programmer's background therefore, I try to avoid use python scripting, however If I there is no other choice I am willing to improve my skills.

My question is as stated on the title, finding a way to improve the performance on OD cost matrix analysis on ArcGIS 10.0.

I use ArcSDE with postgresql and keep my origins and destinations on the database. I have uploaded 200.000 origins in about 75 minutes and it took minutes to load the destinations.

Currently I have waited around half an hour and my work seems to progress around 0%. However It is not crashed yet and according to the status bar it jumps to another location every 5-10 seconds.

As I try to follow "one cup of coffee rule", It seems to be I am doing something very wrong. Do you have any advises to improve performance or at least make sure that my calculations will not crash?

I can only divide origins to smaller parts, I will have to keep my destinations all together.

Kind Regards, Anil.

share|improve this question

2 Answers 2

Normally route calculations are done as 1-to-many, which means it is faster to calculate from one destination to your 200,000 origins, because that only have to be done 1000 times as opposed to 200,000 times. If it is important to you that the routes are calculated from origin to destination, you will have to swap all oneway restrictions temporarily (because a route from A to B isn't always the same as B to A).

Size of the street network also matters a lot. How many records does your underlying street network have?

Right now you are looking at 7.5sec x 200,000 = 25,000 min = 416 hours = 17 days. An awful lot of coffee :-)

share|improve this answer
    
I have around 17 Million records on my street network which is actually whole Turkey. According to your calculation I need to improve the analysis performance around %75, as I have to do this analysis every month in 5 days period upon dataset delivery. –  Anıl Çelik Mar 30 '12 at 11:40
    
Are you sure about this? A street network for the UK with all details are around 4 million records –  Uffe Kousgaard Mar 30 '12 at 15:42
    
Turkey is a big country :) Beside, sea transportation is included on this dataset. –  Anıl Çelik Mar 30 '12 at 19:44
    
Not that big. What is your data source? Anyway, I know it can be done several orders of magnitude faster with the software I sell (www.routeware.dk), but I don't know if NA is really that slow or you are doing something wrong, besides what I have already pointed out. Your street network may be one of them. –  Uffe Kousgaard Mar 31 '12 at 7:01
up vote 0 down vote accepted

If anyone interested, I was only keeping my network dataset and origin file in ArcSDE, I have also imported my destinations to SDE and deleted all unnecessary fields which are stored within my tables yet, I or ArcGIS Network Analyst do not need and I got faster data upload rates now.

Beside, I just realised that I would not use the lines that Network Analyst Rendered, so I have just disabled them before the analysis and the anaysis has been completed even faster. (10 to 20 X faster I believe)

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.