01-14-2015 03:59 AM
I am working on a platform where 3 analytic software can be integrated on a single platform so that the systems can communicate witho each other. How can this happen?
Thanks and regards,
01-14-2015 04:21 AM
By what do you mean communicate? What system are they running on? For a local install, just install all three, then instead of using SAS proprietary datasets, save as CSV, its then available to the other two packages and they can write CSV back (or XML etc.).
01-14-2015 04:54 AM
- you have licensed them all with IML activating the R interface being capable in Enterprise miner also.
- The installation is technical set up with these integration. and you security is set up with fine OS layers opening the XCMD with SAS .
One issue this approach requiring OS layer controls is very tempting for SAS although it is almost mandatory.
01-14-2015 05:25 AM
Thank you RW9, Jaap Karman for your response. I think I am asking the right questions to the right people.All of these are running on a single Linux system .Is there any way we can feed the output from one system as an the input of another system in real time?
Jaap ,I understand what you meant however can you share some more information on the OS Layer Controls?
01-14-2015 06:39 AM
Your question is quite vague. Can you describe your requirements in more detail?
What kind of integration do need (data, calling statistical functions), on what type of interfaces?
System to system, or are there interactive users part of the picture?
What should your system be able to in the end?
01-14-2015 07:55 AM
What do you mean "feed the data in real time?" As you probably know, R works by loading data into memory and then applying methods to it. So you need to supply data in the form of files readable by R. You can't let "look" R into SAS datasets, automatically reacting to changes (esp. since most changes to a SAS dataset are done by completely rewriting it). You can let SAS rewrite a file and R reread it.
01-14-2015 06:35 AM
For R I have http://www.r-project.org/doc/R-FDA.pdf see: chap 7 21 CFR Part 11 Compliance Functionality. All controls should be implemented on OS level. (Your single Linux system).
That access to the system monitoring events etc. Depending on the branch en sensitivity of the data that can be very strict ore a little bit loosely.
For sure you are not going to use and run everything by root. You need root to set up the Linux system harden that and than go for R/SAS/Phyton as toolset.
For the toolset you are likely to do that by an installation key and are the tools up and running close that lock it away.
Normal work you are doing with your personal key all code / data access defined to that one, Beware, there are two different artifact mentioned data and code (business)
1/ for code you are needing release management (Dev Test Prod) ore more variations
2/ for data you are needing some data management (MDM) maybe segregated in Dev Test Prod or more variations.
Having an highpriviledged non personal account defined as the technical owner you can organize this very reliable.
There are several technical complications but most of them are solvable. Are you comfortable with Unix file-systems sudo DAC controls?
01-14-2015 08:38 AM
R can be intergrated with SAS/IML both are in memory matrix operations. SAS/IML Software's Integration with the R Statistical Language
It is extended to enterprise Miner 54806 - Using R in the Open Source Integration Node in SAS® Enterprise Miner(tm)
With proc proto you can call external fucntions. Base SAS(R) 9.4 Procedures Guide, Third Edition Not sure whether Phyton is using the same Call level interface but that seem to me as true 1. Extending Python with C or C++ — Python 3.3.6 documentation
For R usage (matrices) I am seeing they are also switching between Phyton (3 gl functions and a lot high level extensions).
01-15-2015 03:02 AM
Thanks. Learned something new, once again.
Might be useful the next time R comes up as a topic in our organisation.