Flow Classification Optimizations in DPDK
Day01-Session07-SamehGobriel-DPDKUSASummit2016
Day01-Session07-SamehGobriel-DPDKUSASummit2016
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Metrics for Good <strong>Flow</strong> Table Design<br />
Hash value used to <strong>in</strong>dex<br />
<strong>Flow</strong> table<br />
Packet Header<br />
<strong>Flow</strong> Key<br />
H(..)<br />
Payload<br />
Fields of the packet are<br />
used to form a flow Key<br />
Hash function is used to<br />
create a flow table <strong>in</strong>dex<br />
<strong>Flow</strong> Table<br />
Key 1 Action 1 Key 2 Action 2<br />
1. Higher Lookup Rate = Better throughput<br />
& latency<br />
2. Higher Insert Rate = Better <strong>Flow</strong> update<br />
& Table Initialization<br />
3. Efficient Table Utilization = More <strong>Flow</strong>s<br />
Key x Action x Key y Action y Key z Action z<br />
Key N<br />
Action N<br />
Retrieved keys are<br />
matched with <strong>in</strong>put key<br />
Key x<br />
Key y<br />
Key z<br />
<strong>Flow</strong> Key<br />
Match<br />
Action