I have been running Cuffdiff for ~36 hours on a two condition comparison... And it is not finishing, I am using the r3.2x instance. Should I change instances? I seems like it went much faster when I was using the local usegalaxy instance. Thanks!
Allocating more resources (larger cluster/nodes, higher memory allocation) could help. Or you can wait for the job to complete. If it fails for memory (it shouldn't time out on Cloudman with default job resource options), then you will definitely need to move to using an instance/cluster nodes with more dedicated memory.
Thanks! Jen, Galaxy team