Flow Classification Optimizations in DPDK

More documents

Recommendations

Info

Table Load Cuckoo Hashing Performance Benefits Cuckoo Hashing allows for more flows to be inserted in the flow table RTE-hash can be used to support flow table with millions of keys (e.g. 64M – 5 tuple keys) that fits in the CPU cache. 100.00% 90.00% 80.00% 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% Table Load at First Key Insertion Failure Traditional Exact Match Cuckoo Hashing Intel(R) Xeon(R) CPU E5-2699 v4 @ 2.20GHz Hyper-Threading: disabled
Code Snippet for RTE-hash API struct rte_hash *rte_hash_create (const struct rte_hash_parameters *params) int rte_hash_add_key_data (const struct rte_hash *h, const void *key, void *data) int rte_hash_lookup_data (const struct rte_hash *h, const void *key, void **data) int rte_hash_lookup_bulk_data (const struct rte_hash *h, const void **keys, uint32_t num_keys, uint64_t *hit_mask, void *data[]) Reference: http://dpdk.org/doc/api/rte__hash_8h.html
Page 1 and 2: Flow Classification Optimizations i
Page 3 and 4: Memory Flow Classification on Netwo
Page 5 and 6: DPDK Framework Network Functions (C
Page 7: Cuckoo Hashing High Level Overview
Page 11 and 12: Build Flow-Table Insert Performance
Page 13 and 14: Insert Rate (Million ins/sec) Summa
Page 15 and 16: Flow-Table Lookup Performance Optim
Page 17 and 18: Code Snippet for RTE-hash with AVX
Page 19 and 20: POC: Open vSwitch Flow Lookup Packe
Page 21 and 22: Bloom Filter as 1 st Level of Indir
Page 23 and 24: Legal Disclaimers No license (expre

Flow Classification Optimizations in DPDK

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?