• 1 reply
  • Latest Post - ‏2014-08-12T17:57:28Z by macroadster
4 Posts

Pinned topic adaptive MapReduce

‏2013-09-24T22:35:41Z |


I am trying to determine if adaptive MapReduce would be a good option to use instead of the Apache MapReduce. I haven't been able to locate best practices for BigInsights so I'm hoping to get some feedback from the forum community. 





  • macroadster
    5 Posts

    Re: adaptive MapReduce


    Adaptive MapReduce is slightly lower latency than Apache Mapreduce in small clusters (20 nodes or less).  Adaptive MapReduce also offers similar capability as YARN as a resource manager.  Therefore, if you are looking for parallel computing and not necessarily mapreduce, Adaptive MapReduce offers alternative methods to spawn off parallel tasks more efficiently.  Unlike YARN that Application needs to implment application logic to negoicate with resource manager.  Adaptive MapReduce can configure resource through ConsumerGroups and Resource Groups using properties.  Therefore, there is no single point of failure to track application state.  It is a simplier design than YARN.

    Adaptive MapReduce enables to run existing IBM software more seamlessly, where Apache MapReduce works better with pure open source software.

    Hope this helps.