10Gbps Scalable Flow Generation and Per-flow Control with Hierarchical Flow Aggregation Decomposition Using IXP2800 Network Processors

摘要

With the growth and development of the Internet, understanding the composition and dynamics of network traffic is greatly important for network engineering, planning, design, and attack detection. This paper proposes 10Gbps dynamic flow monitoring and control using IXP2800 network processors with flow dynamic rate-based flow aggregation that ensures scalability of the system. We propose fast and scalable 6-tuple TCAM-based classification, which provides three levels of aggregation and flow control that includes flow metering, marking, queuing and scheduling. Using flow control allows network operators to adjust QoS-level dynamically and restrict malicious activity in the network. The proposed implementation design is based on Radisys ATCA-7010 processing module, containing two IXP2800 network processors and TCAM module for highspeed classifications. Flow information collector can be either implemented on another processing blade, or on external computer. The proposed architecture supports up to 125,000 flows in total.