4 weeks ago - last edited 4 weeks ago
I would suggest the following considerations:
1 – how many people need to run the software?
If multiple users need the software, maintaining multiple individual installs is really inconvenient since it interferes with true project sharing and adds installation/configuration maintenance challenges. Each machine will need to be configured individually rather than needing only a simple thin client on the workstations that connect to the server. You will have to move data and projects back and forth to share work which is inefficient. On a server, there is one copy of each project and each user can access any available diagram at the same time. Should one user be out sick, the other users can continue to work.
2 – what you expect to do with your machine while SAS EM is running?
If you anticipate using your workstation for other work while SAS Enterprise Miner is running, you will have far less computing power/memory available in the workstation installation since the software will be using your physical and virtual RAM while it is running. As a result, your other tasks might take far longer to accomplish making your other work less efficient. This is especially true if you are dealing with substantial data sets. If a server is doing the work, you have virtually all of the power and disk space on your workstation available to work on other tasks.
3 – where does your data live?
If your data is in small datasets that can be easily moved, there isn’t really any difficulty although it is inconvenient to have to move it back and forth; however, if you are accessing remote data (geographically or physically removed from the compute tier), you will end up wasting a lot of I/O as your workstation reads and writes data from the network location. SAS Enterprise Miner uses the directory where the data resides to write out temporary data sets that are used during processing. You will need to keep accessing the data remotely which can greatly slow processing. If you keep the data on your workstation, processing will be faster but you are severely limited in the available space which will almost limit performance and will compound the issue noted in number 2 above.
Hope this helps!