Postgres by default has a configuration that is not well suited to loading or retrieving large volumes of data (for example gene expression, or large sets of clinical observations).
In practice, a number of optimizations have been found to help the performance and scalability of data loading and of retrieval for export and analysis by tranSMART.
Performance tuning parameters can be calculated online with http://pgtune.leopard.in.ua/ using the physical memory installed on your postgres server, and a figure of 50 connections. Postgres configuration is set in file postgresql.conf in the main postgres directory.
Postgres can be relocated to another disk (for example in AWS instances) by stopping the postgresql server and moving the pgsql directory (this contains the postgres and database files, by default in the data subdirectory) and adding a symlink before restarting the postgresql server.
The maximum number of connections should not be too small. Even for a one-user server there can be many connections at one time for some queries in loading expression data and in the Analyze tab when retrieving concept count values for the tree. In such cases ETL can fail - a connections error should appear in the postgres log file.