Frequent  Itemset  Mining  (FIM)  is one of the  most well  known  techniques   to  extract   knowledge  from  data.   The combinatorial  explosion  of  FIM  methods  become  even  more problematic   when  they  are  applied  to  Big  Data.  Fortunately, recent improvements in the field of parallel  programming already provide  good tools to  tackle  this  problem.  However,  these  tools come  with  their  own  technical  challenges,  e.g.  balanced   data distribution  and  inter-communication  costs.  In  this  paper, we investigate the applicability  of FIM techniques on the MapReduce  platform.   We  introduce   two  new   methods   for   mining   large datasets: Dist-Eclat  focuses on speed while BigFIM is optimized to run  on really large datasets.  In our  experiments  we show the scalability  of our methods.