mapreduce - How Hadoop calculate physical memory and virutal memory during a job execution -
i have few queries related counters used in hadoop display memory usage. map reduce job executed on cluster gives me below menitoned counter values. input file used in kbs, these counter shows 35gb , 420 gb usage. physical_memory_bytes=35110662144 virtual_memory_bytes=420121841664
for different job on same input file shows 309 mb (physical) , 3g(vitual) usage
physical_memory_bytes=309526528 virtual_memory_bytes=3435827200
first job more cpu intensive other , creates more objects other 1 still usage shown seems high. wanted know how memory usage calculated. tried going through posts , gave on view on below link seems requirement task describing these variables (https://issues.apache.org/jira/i#browse/mapreduce-1218 ) couldnt find how these calculated. gives me idea on how these values passed job tracker,but no information on how these determined. if 1 give insight on helpfull.
you can find few references here , here. second link in particular map , reducer job , how slots decided based on memory allocations. happy learning
Comments
Post a Comment