Scalable Many-Field Packet Classification using Multidimensional-Cutting Via Selective Bit-Concatenation

摘要: OpenFlow Switch in Software-Defined Networking (SDN) has changed packet classification from standard 5-tuple to arbitrary many-field. The growing number of fields a rule and the increasing rules ruleset poses great challenges for terms performance, storage, update cost. In this paper, we design two-stage system address those issues by exploiting sparsity independence. A is examined offline with proposed matrices find representative bits different field rule. We leverage concatenate them as sample values divide into several subsets spaces. Each subset given unique each space. only affects related addresses. pre-filtering stage comes out highly intersecting candidate spaces full match process. Out throughput 356 MPPS 1K 15-field 213 100K when using single NVIDIA K20C GPU card.