Receive-side Scaling circuit for multi-core processor dynamic grouping management

Abstract

The invention discloses a receive-side Scaling circuit for multi-core processor dynamic grouping management. The circuit includes a TCP / IP head extraction module, a Hash _ Function module, a Hash _ Mask module, an indirect table query module, a CPU _ Result calculation module and an RSS register module; the TCP / IP head extraction module is linked with an MAC module for completing encapsulation, deblocking, sending and receiving functions of data frames; the CPU _ Result calculation module is linked with an RQP module for allocating the data frames to different receiving arrays; the RSS register module is linked with the TCP / IP head extraction module, the Hash _ Function module, the Hash _ Mask module, the indirect table query module and the CPU _ Result calculation module separately. The receive-side Scaling circuit effectively achieves dynamic grouping management of CPU cores in a multi-core processor system so that the CPU cores with high load can be shut down in real time, and the processing performance is dynamically adjusted during data reception to make the network throughput rate reach the maximum value.

Description

technical field [0001] The invention belongs to the technical field of computer networks, and in particular relates to a Receive-side Scaling circuit for dynamic group management of multi-core processors, especially suitable for multi-processor systems that need to balance network data throughput according to system load and network processing performance, such as servers , measurement and control system central computer and Linux multi-core platform, etc. Background technique [0002] At present, Receive-side Scaling (RSS, receiving-side regulation) technology is used in mainstream Ethernet controllers, mainly because the RSS technology is relatively mature and has good software compatibility. [0003] The existing RSS technology analyzes and processes the received data message, resolves the IP source address and destination address (IPV4 or IPV6), TCP source port and destination port in the message, and then calculates the hash_type and hash_key values hash_rusult, accord...

AI Technical Summary

Problems solved by technology

[0005] 1. It is impossible to dynamically adjust the CPU core according to the capacity of the network load;

[0006] 2. The group management of CPU cores cannot be realized, so that the CPU cores with higher load capacity are shut down

[0008] 1. The hash value obtained by the Toeplita_Hash function determines which queue the message is assigned to, but the initialization value of Toeplita_Hash is configured after power-on and cannot be modified, so the dynamic balance processing of data packets cannot be realized;

[0009] 2. When the load of some CPU cores reaches the upper limit, it cannot be closed dynamically, resulting in the discarding of data packets

Images