What are you actually trying to achieve? Your question is extremely vague.
Do you just want a big collection of drug-like molecule? Grab GDB-17
Do you want a curated source of biologically interesting molecules? Go to ChEBI
Do you want small molecules with annotated activity? Try ChEMBL
Do you want small molecules bound to a protein? Try PDB
For docking, just go to any literature search engine and search for small molecule docking. You will get reviews, algorithms and papers describing software. There is plenty of literature, so start reading.
Edit: After several comments, it seems the interest lies in identifying any compound which may inhibit the activity of the given receptor. For this purpose I would suggest using ZINC, which also has vendor information.
As you suggested, you will find a smaller but usefuly target mapped compound and drug set at http://www.guidetopharmacology.org/
You should probably start docking with PDB structures anyway so you can get a small set from the above by using the PubChem query
https://www.ncbi.nlm.nih.gov/pccompound?term=%28%231%29%20AND%20pccompound_structure[filt]%20AND%20%22IUPHAR/BPS%20Guide%20to%20PHARMACOLOGY%22[SourceName]&cmd=DetailsSearch = 872
the protein sequences i am planning to work with , doesnot have any available structure. so i am developing a molecular modeling based study first then the screening of molecules to dock with the predicted model. So what kind of molecular databases suit to this?
my goal of work is to determine few novel molecules which can inhibit the activity of the receptor of interest..( determining potential compounds that can be taken for clinical trial), that need not be FDA approved or already known drugs..the study i would like to follow is molecular modeling based ( sorry for not to add modeling in earlier ).
thanks for the suggestions!