Sharing data between separate SAS instances ?

[Chris_R]

Looking for a solution to a problem which I have been presented with.

We currently have two separate SAS installs at our site - both 9.4M6, one is used for BASE SAS code and the other is a full stack including web services, SNA, DI amongst others.

We have a user group that uses both servers and would like access to several (large > 100GB sas7bdat size) files from both systems. We normally transfer (rsync replication) over smaller files but larger data sets would be problematic to transfer in the window we have when the servers are fairly idle overnight.

In the past we used to use a hosted share on one of the servers, but this also proved problematic, often leading to locked files, especially on pre-assigned libs and then delays on startup of SAS sessions waiting for timeouts to happen. It also means that we had to co-ordinate the downtime of both instances so that no jobs using these shared tables are in process of being written to from the other server or risk loss of data.

Looking at SAS/SHARE - would this be a possible solution to transferring the files over wholesale each evening and also allow us to attribute file locks effectively between the two separate platforms?

The other plan I am investigating is a dedicated file host that we can have both system independently attach to with the shared data sets housed on it, but again I am not sure that SAS will be able to "know" that a file is locked for single access or being written to by the other server.

[gwootton]

SAS/SHARE would make the data available through a connection to the SHARE server. So your users would connect to the library as a connection to the share server rather than opening the file locally.

[Sajid01]

100 GB+ of data in a sas dataset is a fit case to be stored in and RDBMS system.
Accessissues from multiple servers will automatically be solved.

[Kurt_Bremser]

Will the dataset(s) in question be written from both servers, or only from one?

[Patrick]

Besides of SAS/Share being single threaded it still wouldn't really resolve the issues with SAS/Share being down - and it needs to run on one of the SAS Servers.

If you want to/must stick with data replication then just as a thought (and not tested): If you would use the SPDE engine for your >100GB tables then only a portion of the partitions might change on a daily basis so using rsync could still work as it will only copy the partitions that changed.

[MargaretC]

Can you put a shared file system that both SAS infrastructures can access?  That is all that is needed for READ only access to the files. 

SAS/SHARE should only be used if you are needed for both SAS infrastructures to update the files at the same time.