In the benchmarks, DNS_QPGC_ALL was trying to hard to cleanup QP and this was slowing down QP too much. Use DNS_QPGC_MAYBE instead that we are going to use anyway for more realistic load - this also shows the memory usage matching the real loads.
In the benchmarks, DNS_QPGC_ALL was trying to hard to cleanup QP and this was slowing down QP too much. Use DNS_QPGC_MAYBE instead that we are going to use anyway for more realistic load - this also shows the memory usage matching the real loads.