We recently acquired the ArcInfo suite for a varitey of uses in our organization and we're having some struggles with performance.
Here's a short description from my end user on how he is using the applicaiton:
[INDENT]So far I have mainly been combining TIGER files from the Census bureau and shapefiles from the City of Chicago and then joining those files with data from the 5 year 2009 ACS and the 2010 Census. Once that data is joined, I either change the symbology to visually depict the information or I go forward with the steps to interpolate data and create heatmaps of the demographics
When I use census tracts the computer struggles sometimes, but it is a bearable wait time between operations. Unfortunately the census tract level data means making a lot of generalizations about swathes of people in the city. There are 1318 Census Tracts in Cook county with the majority of those tracts in the city. There are roughly 4 block group pieces in each Census tract with 3992 block groups in cook county and in each block group there are 25 blocks.
The computer starts to really slow down when it needs to render 99,042 blocks and then crawls when I try to do analysis operations on the block level data or render multiple maps for presentations. I do need block level shapefiles for upcoming strategy sessions, and on top of that I sometimes need two or three sets of the block level shapefiles on the screen at once in the layout view to adjust them for presentations.
I�??d love to eventually perform operations with our membership data as a whole, but it takes 25 minutes to geocode the membership data and then about a minute to render completely each time the map is altered. Even then, our membership spans 7 counties, so if I need to pull in block or block group data from other areas of Illinois to cover our membership I get closer to 300,000 shapefiles. [/INDENT]
I also have heard reports of the operations crashing randomly while processing.
We have the suite installed on a Dell OptiPlex 380, outfitted with an Intel Core 2 Duo E7500 (Wolfdale), 4GB of memory and an Nvidia Quadro 400 graphics accellerator. It appears as though there is plenty of free memory while performing these tasks.
I've got the go-ahead to order whatever type of machine we need to get some reasonable performance out of the application, but my struggle comes with the fact that while the application is working at full throttle, it's only using one processing core. The wording about multicore support on the ESRI website is ambiguous to say the least, but when you really drill down it basically states that applications are not multithreaded. Multiple cores are supported in the "One for the app, and the rest to do whatever else the machine needs to do.." scenario. Our sales rep echoed this sentiment, and states other customers have been asking for the same thing. I was told it's coming, it's just not in this release of the software.
I can acquire an I7-2600 load it up with RAM and a decent video accellerator, but if the application is going to only use one processing core, I'm not sure I'm going to see much of a performance increase. I know the Sandy Bridge processors are more efficient in that they can get more work done per clock cycle, but I don't know how much of an improvement we are going to see here.
Any advice? Thanks in advance!
