Event-Driven Approach for Flow-to-Core Mapping by NICs in Multicore Systems

In exploiting multiple cores for packet processing, steering incoming packets to the same core on which its application resides is important for the performance issue. One of the key components for this is flow-to-core mapping table, which is used by NIC to steer incoming packets to right cores. To acquire the information required for flow-to-core table, we propose an event-driven approach unlike existing technologies, which take per-packet-based approach and thus cause non-negligible overhead. In addition, exploiting standard Linux utility to manipulate flow-to-core table of NIC, we realize NIC-independent approach unlike existing technologies, which require specific NICs.

[1]  Wenji Wu,et al.  Why Can Some Advanced Ethernet NICs Cause Packet Reordering? , 2011, IEEE Communications Letters.

[2]  Dipak Ghosal,et al.  Cache-aware affinitization on commodity multicores for high-speed network flows , 2012, 2012 ACM/IEEE Symposium on Architectures for Networking and Communications Systems (ANCS).

[3]  Hyun-Wook Jin,et al.  MiAMI: Multi-core Aware Processor Affinity for TCP/IP over Multiple Network Interfaces , 2009, 2009 17th IEEE Symposium on High Performance Interconnects.

[4]  Robert Tappan Morris,et al.  Improving network connection locality on multicore systems , 2012, EuroSys '12.

[5]  Wenji Wu,et al.  A Transport-Friendly NIC for Multicore/Multiprocessor Systems , 2012, IEEE Transactions on Parallel and Distributed Systems.