Hi All,
I have two questions for which I will need to provide some background. This is
not my field of expertise, so I'm not sure if the terminology I use is
necessarily the 'standard'. I'm working off a UNIX system. Please email me
direcly, and I will summarize the replies I receive. Thanks in advance.
We are in the process of setting up an analytical data warehouse. All of the
data loading (from multiple sources), and analysis is carried out in SAS. The
end users are provided with access to the staged data after quite a bit of
work (transforms, summations, normalization, etc.) has been carried out on
the input data. All of the staged data is kept in one partition where only
power users have access rights. All other users will be provided access to
another partition with bot SAS and Oracle datasets. There are about 70
million records, this work will be done on a monthly basis, data is stored
for the latest 3 months, so speed and space considerations become relevant.
- Question 1 (SAS datasets) ----
On the user partition, we can provide end users access to the SAS datasets in
the poweruser partition by creating
a) copies of the data - not an option due to space considerations
b) soft links - this is a viable option but the final datasets in the power
user partition are very unpolished i.e., they contain extra variables (not
necessarily the same set for each month) that are not defined in documentation
(or in Oracle), no labels. We'd prefer to have a perfect match between the SAS
and ORACLE data copies.
c) sas data views - this is the option we're looking into at the moment,
because at this stage we can label the variables, add/drop variables, etc.
basically clean up before presentation.
Question - Does anyone have experience with SAS data VIEWs? I've been playing
around with them a little bit to check speed of access, etc. and am getting
very mixed results. Many of our in-house macros do not function when used
directly on a VIEW (eg. the 'point' option in a datastep.) What are the other
(dis)advantages I should be aware of with this approach? Any other factors I
should consider? I would very much appreciate hearing from people with
experience with data VIEWs.
- Question 2 (SAS to ORACLE Connectivity) ---- We're loading the final
staged version of the data into ORACLE tables using SAS ACCESS. Once again,
the datasets are large. The process is cumbersome and is taking way too long.
We're getting load times of up to 26 hours on one of our larger tables. I get
the feeling that SAS Access is a white elephant, it's making it convenient by
allowing us to keep working in SAS but charging us heavily with it's
inefficiency. I would really appreciate hearing from others with experience
using SAS Access. Are there any other tools out there one could use as a
replacement for SAS ACCESS? Would we be better off creating a CDF or other
ASCII file to download into ORACLE? Anyone with experience on using SAS
ACCESS on VIEWs?
Thank you all in advance. Please post replies direcly to me and I can post a
summary later, or send out copies of replies to others with an interest.
Sincerly,
Samir Mishra.
-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
Received on Thu Apr 29 1999 - 10:09:45 CDT
