I have a TDM subset/masking plan that will (usually) not run. Sometimes I get an error message related to cache ("unable to allocate a cache block" type error), but most of the time the session log just ends abruptly. Often the last line in the log is related to a substitution cache. I have tried many different combinations of dictionary sizes and TDM Advanced Plan Settings for Cache Size with mixed results. What does TDM Cache Size really control? Is it the amount of memory for each dictionary cache or the total amount of memory split across all of the dictionary caches? Is it on top of the Power Center setting for Maximum Memory Allowed for Auto Memory setting (5%) or in addition to? We have increased the RAM on the Informatica server to 64GB and I have moved the cache directory to a drive with 99GB free. I am not short on either RAM or disk space. What else could influence the space allowed for substitution cache? I have run the plan successfully once or twice by selecting only 1 record from the driving table. The entity has 188 DB2i5OS tables and the source/target connections are Power Exchange. The one large dictionary has 1.5 million rows, but I have cut that down to .5 million for different trial runs. Any suggestions?