Accumulators corresponding to bins in memory

A system comprising:

• one or more hardware processors;
• a plurality of processing entities comprising a first processing entity and a second processing entity executable on the one or more hardware processors;
• a memory to store data arranged in a plurality of bins associated with respective key values of a key;
• and
• a cache to store first cached data elements for respective first accumulators, and second cached data elements for respective second accumulators, wherein the first cached data elements are part of a first accumulator structure associated with the first processing entity, and the second cached data elements are part of a second accumulator structure associated with the second processing entity, wherein the first and second cached data elements are updatable to represent occurrences of the respective key values of the key, wherein each accumulator of the first and second accumulators corresponds to a different bin of the plurality of bins, and each cached data element of the first and second cached data elements has a range that is less than a range of a corresponding bin of the plurality of bins,
• wherein the plurality of processing entities are to apply data analytics on input data records in parallel with one another, and the plurality of processing entities are to use respective accumulator structures that are private to respective processing entities of the plurality of processing entities so that the plurality of processing entities do not contend for access of any of the accumulator structures, wherein the accumulator structures include the first accumulator structure and the second accumulator structure, and
• wherein responsive to a value of a given cached data element as updated by a given accumulator of the first and second accumulators satisfying a criterion, a respective processing entity of the plurality of processing entities is to cause an aggregation of the value of the given cached data element with a bin value in a respective bin of the plurality of bins.

A non-transitory machine-readable storage medium comprising instructions that upon execution cause a system to:

• store a shared data structure in a memory, the shared data structure being shared by a plurality of processing entities and being arranged in a plurality of bins associated with respective key values of a key;
• store, in a cache, accumulator structures for respective processing entities of the plurality of processing entities, wherein each accumulator structure of the accumulator structures includes cached data elements for respective accumulators, the cached data elements being updatable to represent occurrences of respective key values of the key, wherein each accumulator of the accumulators corresponds to a different bin of the plurality of bins in the shared data structure;
• and
• responsive to a value of a given cached data element as updated by a given accumulator in a first accumulator structure of the accumulator structures satisfying a criterion, initiate, by a first processing entity of the plurality of processing entities, an addition of the value of the given cached data element to a respective bin of the plurality of bins in the memory without using any atomic write operation.

A method of a system comprising a hardware processor, comprising:

• storing a shared data structure in a memory, the shared data structure being shared by a plurality of processing entities and being arranged in a plurality of bins associated with respective key values of a key;
• store, in a cache, accumulator structures for respective processing entities of the plurality of processing entities, wherein each accumulator structure of the accumulator structures includes cached data elements for respective accumulators, the cached data elements being updatable to represent occurrences of respective key values of the key, wherein each accumulator of the accumulators corresponds to a different bin of the plurality of bins in the shared data structure;
• and
• responsive to a value of a given cached data element as updated by a given accumulator in a first accumulator structure of the accumulator structures satisfying a criterion:
• determine, by a first processing entity of the plurality of processing entities, which processing entity is an owner processing entity of a portion of the shared data structure into which the value of the given cached data element is to be added,
• in response to a determination that the owner processing entity is not the first processing entity, transfer the value of the given cached data element to the owner processing entity, and
• add, by the owner processing entity, the value of the given cached data element to a respective bin of the plurality of bins in the memory.