Finding related structure
     
The first thing, of course, is to visit PDB and type in the name of the protein in the search window. Download the structure(s) that you like. Multiple structures might represent different conformations of your protein, or show it bound to different ligands. You can upload them into Cube server later to have the conservation/specialization score mapped onto them.
     
  If, for some reason, this does not work out (name resolution is a somewhat tricky business), while still on the PDB homepage, try pasting in the sequence in the search window (note you need to select the "sequence" tab above it).
     
  A similar option is offered by NCBI BLAST site just make sure to choose PDB as the database against which to do the BLAST search.
     
If this still does not work, or you are getting a structure for only one portion of your protein, it is possible that nothing similar was ever crystallized. Sometimes it is possible to model the structure of (a part of) your protein.
UNIPROT pages usually contain several links to modeling servers and databases. Here is an arbitrary example - once on the page corresponding to your protein, scroll down to the "Cross-references" section. You should see a small menu for the 3D structure databses, like the one here, on the left.  
     
Still no results? Consider the possibility that your protein is intrinsically disordered. Disopred is one of the websites that will give you a prediction, an estimate of how likely it it is that your protein, or regions thereof, are disordered. In this case we'll have to stick to 1D analysis. Cube will produce an xls spreadsheet and a png map for that purpose.