Need help in caching and persist

[Jeetendra Gangele <ga...@gmail.com>]

Hi

I have a job in which I am doing 2-3 transformation of my data.

Lets say, step#1 i am transforming RDD<Object> to RDD<Object,List<Object>>
using mapTo pair function. this  will result in 30 GB data .
Should i cache or persist this data or leave it later for conversion?

in next step I will again transform the step#1 RDD into another RDD of
<Object ,list<FeatureVectorForMLLIB>>.. should I cache or persist it? this
will be much lesser in size compare to step#1 RDD may be around  2 GB

in next step I will run the the foreach and ask my mode to predict? the out
for each combination from Step#2 RDD.

Overall what are practices when to cache or persist

regards
jeetendra