Is the intention for this to be a grid environment where the compute tier hosts have a shared configuration directory, and compute load is balanced among the compute hosts (i.e. connecting to SASApp Workspace Server results in a Workspace Server session starting on host 1, 2 3 or 4)? Or would each compute tier be a separate application server (i.e. SASApp on host 1, SASApp2 on host 2, etc)? In a grid environment you can control workload placement based on resource requirement settings or queues. If each host has its own application server context, the user could select the context they want and thereby the associated host, but no load balancing would be present. If a user was connected to one host and needed libraries that were local to another the path would need to be shared between them or some service would need to make them available remotely (i.e. SAS/SHARE, SAS/CONNECT Remote Library).
-- Greg Wootton | Principal Systems Technical Support Engineer
What OS is your SAS servers running on? Setting up folder shares between your SAS compute servers for the SAS libraries that are being shared is another possible approach as long as they are all running the same OS.
The same OS, or only compatible data representation? We can read and write big-endian Solaris data sets from little-endian Windows through a mounted drive, but it uses Cross Environment Data Access and might not be efficient.
In a grid environment, some shared file systems work better than others. There's a SAS paper about it. Expensive file systems tend to handle voluminous simultaneous updates from multiple systems better than the cheaper ones do.
You have two questions. Q1 : How the users can connect to a specific Compute? (permissions defined in the Management Console?)
The users connect through the metadata tier. Q2 :If a user is connected to Comp2, is it possible for him to access to libraries on Comp1 ? Yes.
For a young and up coming administrator, with the questions you have asked in mind, I suggest that you read the following references. They will server as a good starting point (in addition to the SAS documentation) https://www.lexjansen.com/nesug/nesug11/ma/ma10.pdf