Scalable TCP Session Monitoring with Symmetric Receive-side Scaling
نویسندگان
چکیده
Receive-side scaling (RSS) is a technique that stores the arriving IP packets in the same flow into the same hardware queue of a modern network interface card (NIC). It allows scalable processing of the received packets by allowing exclusive access to the NIC queues by each CPU core. This removes the lock contention when accessing the NIC queue, and it allows concurrent access to different queues by multiple CPU cores. One problem with the existing RSS mechanism, however, is that it maps the IP packets in the same TCP connection to different NIC queues depending on the direction of the packets. In this paper, we present symmetric RSS, which allows mapping the packets in the same TCP connection to the same NIC queue regardless of whether they are upstream or downstream. The basic idea is to manipulate the RSS seeds such that the RSS hashes of the IP packets in the reverse direction to take on the same values with those in the forward direction. Since RSS hash calculation is done in the NIC hardware, it does not consume extra CPU cycles and promises high performance.
منابع مشابه
Analysis of Scalable TCP in the presence of Markovian Losses
In high speed networks, the standard TCP’s AIMD algorithm was observed to be inefficient in utilizing the link capacity. As a result of which several proposals such as High-Speed TCP, FAST and Scalable TCP, were put forward. In contrast to the additive increase multiplicative decrease algorithm used in the standard TCP, Scalable TCP uses a multiplicative increase multiplicative decrease (MIMD) ...
متن کاملA Simulation Study of Scalable TCP and HighSpeed TCP in Geostationary Satellite Networks
This paper investigates the performance of two TCP enhancements (i.e., Scalable TCP and HighSpeed TCP), recently proposed for high-speed backbone networks with a very large bandwidth-delay product, in a geostationary satellite environment. Both persistent and elastic traffic patterns are considered, performance being evaluated in terms of TCP throughput and mean session delay, respectively. The...
متن کاملOptimizing TCP Receive Performance
The performance of receive side TCP processing has traditionally been dominated by the cost of the ‘per-byte’ operations, such as data copying and checksumming. We show that architectural trends in modern processors, in particular aggressive prefetching, have resulted in a fundamental shift in the relative overheads of per-byte and per-packet operations in TCP receive processing, making per-pac...
متن کاملFuzzy Logic-Based Round Trip Time Scaling and Scheduling in High Speed TCP Stacks
Network management and control is a complex problem that requires intelligent, control methodologies to obtain satisfactory performance. Round trip time (RTT) scaling mechanism has been introduced for changing congestion window and to perform congestion control satisfactorily in all circumstances. This paper presents a fuzzy RTT scaling (FRTTS) scheme that performs RTT scaling and RTT schedulin...
متن کاملMainstream NUMA and the TCP/IP stack
Historically, approaches to achieving scalable networking focused on reducing the host processing requirements associated with TCP/IP interrupts. In the many‐core era, approaches such as interrupt moderation, jumbo frames, and the TCP Offload Engine prove inadequate because the modest increases in the processing speeds forecast for host computers are not keeping pace with improvements in networ...
متن کامل