-
dbis
-
Publications
-
2012
-
Efficient Frequent Item Counting in Multi-Core Hardware
(P)VLDB 2012 Reviews
number of cores is increased beyond 2. Is the choice of pipelined parallelism coming back to bite you? D5: The presentation can be improved in a number of places: -- Unclear sentence in Section 1: To date, [...] a) The machines used are comparable. b) When [20] explores a Zipf distribution with skew parameter 0.5, the bandwidth is between 100 million records/sec (large domain cardinality) and 1100 million records/sec [...] Given the performance results of [20], the dismissal of hash-based techniques is also questionable. D5. Similarly, the authors claim that the combination step is a bottleneck. This may be true for the S …