“…Comprehensive studies across many contemporary Big Data systems [18] confirm that these overheads lead to significantly reduced scalability-e.g., applications crash with OutOfMemoryError, although the size of the processed dataset is much smaller than the heap size-as well as exceedingly high memory management costs-e.g., the GC time accounts for up to 50% of the overall execution time. Despite the many optimizations [6, 7, 16, 19, 21, 23-25, 33, 38, 41, 45, 48, 49, 52, 54-57, 60, 61, 72, 73, 76] from various research communities, poor performance inherent with the managed runtime remains a serious problem that can devaluate these domain-specific optimization techniques.…”