The minimum computer requirements to run the desktop version of SAS EM/TM are too low, in my humble opinion.
http://support.sas.com/documentation/cdl/en/emag/65360/PDF/default/emag.pdf suggests 2GB available RAM and multiple processor cores. This is what I currently run and it takes way to long to run processes.
The minimum requirements are not the 'practical' requirements. In practice, what would your ideal computer specs be to run this software?
In practice, the minimum requirements are exactly that -- the minimum requirements. The software runs but it is not running fast enough for you to be satisfied. It is important in these situations to assess several possible bottlenecks which might be impacting your processing. If your limiting factor is the RAM on your machine, it does not help to consider improving your network or the moving your data to allow faster access from the compute server.
Here is a list of some things to consider:
1. System RAM -- You mentioned your system only has 2 gB of RAM for SAS Enterprise Miner but the software provides several memory intensive methods. Running other applications will cut into this RAM. Increasing your RAM will all but certainly help.
2. Disk space -- How full is your disk? Your ability to allocate extra space via virtual memory is hampered if you are limited by disk space.
3. Disk speed -- How fast is the data access from your machine? If you are accessing data in a remote location or trying to access it off a separate drive, the I/O for data access comes into question as well as the
4. Network speed -- How quickly does data move across your network? Wireless movement of data will be slower (typically) than wired connections. Connections to data located in other buildings or geographical locations can slow down processing.
5. Competition for Resources -- What other people are accessing the data? What other applications are running on the same server? What other bandwidth is being consumed by competing traffic? I have seen systems brought to their knees when everyone started surfing at lunchtime.
6. Data -- Has the data been defined efficiently? Are there unnecessarily wide fields and/or unnecessary variables in the data set? How long are the field names/levels? Applications that export text data might also be exporting formats forcing simple variables (e.g. Yes/No or 1/0 variables) to have a width of 200 or more. SAS must allow for the entire formatted length of the variable so that these unnecessarily wide variables consume far more memory than is needed.
7. Modeling Approach -- If you are not able to improve your performance satisfactorily from taking steps implied above, perhaps you can sample your data. One approach to performance is to simplify the data and/or modeling approach and/or to choose a sampling strategy if one hasn't been implemented.
If you post the specifications of your machine, explain the nature of the data (including variable lenghts), describe the analyses being done or share a screenshot of your flow, and respond to the questions above, it is possible that I could hazard a better guess where your particular opportunities for improved performance lie but I suspect your RAM is the limiting factor at this point.
I hope this helps!
Doug
In practice, the minimum requirements are exactly that -- the minimum requirements. The software runs but it is not running fast enough for you to be satisfied. It is important in these situations to assess several possible bottlenecks which might be impacting your processing. If your limiting factor is the RAM on your machine, it does not help to consider improving your network or the moving your data to allow faster access from the compute server.
Here is a list of some things to consider:
1. System RAM -- You mentioned your system only has 2 gB of RAM for SAS Enterprise Miner but the software provides several memory intensive methods. Running other applications will cut into this RAM. Increasing your RAM will all but certainly help.
2. Disk space -- How full is your disk? Your ability to allocate extra space via virtual memory is hampered if you are limited by disk space.
3. Disk speed -- How fast is the data access from your machine? If you are accessing data in a remote location or trying to access it off a separate drive, the I/O for data access comes into question as well as the
4. Network speed -- How quickly does data move across your network? Wireless movement of data will be slower (typically) than wired connections. Connections to data located in other buildings or geographical locations can slow down processing.
5. Competition for Resources -- What other people are accessing the data? What other applications are running on the same server? What other bandwidth is being consumed by competing traffic? I have seen systems brought to their knees when everyone started surfing at lunchtime.
6. Data -- Has the data been defined efficiently? Are there unnecessarily wide fields and/or unnecessary variables in the data set? How long are the field names/levels? Applications that export text data might also be exporting formats forcing simple variables (e.g. Yes/No or 1/0 variables) to have a width of 200 or more. SAS must allow for the entire formatted length of the variable so that these unnecessarily wide variables consume far more memory than is needed.
7. Modeling Approach -- If you are not able to improve your performance satisfactorily from taking steps implied above, perhaps you can sample your data. One approach to performance is to simplify the data and/or modeling approach and/or to choose a sampling strategy if one hasn't been implemented.
If you post the specifications of your machine, explain the nature of the data (including variable lenghts), describe the analyses being done or share a screenshot of your flow, and respond to the questions above, it is possible that I could hazard a better guess where your particular opportunities for improved performance lie but I suspect your RAM is the limiting factor at this point.
I hope this helps!
Doug
Build your skills. Make connections. Enjoy creative freedom. Maybe change the world. Registration is now open through August 30th. Visit the SAS Hackathon homepage.
Register today!Use this tutorial as a handy guide to weigh the pros and cons of these commonly used machine learning algorithms.
Find more tutorials on the SAS Users YouTube channel.