1.研究雙向鏈結串列
2.Study "Network measurement for 100gbps link using multicore processors"
https://www.evernote.com/l/ANvCc5W3DulMhbeBUNGtqnFy1Xzd7YJMXyk
Highlight :
DPDK, x86 multicore processors,
CountMin sketch, Simple hash table, Revise sketch,
Separate / Shared (where to put sketch table or Hash table)(In L2 or L3 cache)
Receive Side Scalable(multi-queues) / Non-RSS SYNC(only one queue, wait all slave core finish) / Non-RSS ASYNC (only one queue, use ring buffer)
Results :
Best performance >> Simple hash table + Separate + Receive Side Scalable(RSS need NIC support)