huang zuxing blog @ home the quieter you become, the more you are able to hear.


mentor graphic calibre drc job run time



Our cluster daily job throughput

After merged a few pools into a single cluster, daily job number throughput, average is around 150K.


Filed under: eda, IT Comments Off


In case EDA people search those key words,  I might help to answer the questions.

Since 2006, we use below concepts:

free_slot = total compute node cpu number in cluster

calibre_slot = calibre remote cpu number

tflex_slot =  brion tflex remote leaf cpu number

$oreng file , A signal in calibre host runfile

lsf_linux, tapeout_tflex , our scripts to submit mentor graphic calibre jobs and tflex jobs.

with this kind of concept, we can make sure EDA applications which use master and slave compute concept have enough cpu/licenses to run jobs.


eda-cost-savers, doc from teraproc

Original doc on

Ten ways to reduce the cost of your EDA infrastructure