High-Bandwidth Network Memory System Through Virtual Pipelines

Abstract

Abstract-As network bandwidth increases, designing an effective memory system for network processors becomes a significant challenge. The size of the routing tables, the complexity of the packet classification rules, and the amount of packet buffering required all continue to grow at a staggering rate. Simply relying on large, fast SRAMs alone is not likely to be scalable or cost-effective. Instead, trends point to the use of low-cost commodity DRAM devices as a means to deliver the worst-case memory performance that network data-plane algorithms demand. While DRAMs can deliver a great deal of throughput, the problem is that memory banking significantly complicates the worst-case analysis, and specialized algorithms are needed to ensure that specific types of access patterns are conflict-free.We introduce virtually pipelined memory, an architectural technique that efficiently supports high bandwidth, uniform latency memory accesses, and high-confidence throughput even under adversarial conditions. Virtual pipelining provides a simple-to-analyze programming model of a deep pipeline (deterministic latencies) with a completely different physical implementation (a memory system with banks and probabilistic mapping). This allows designers to effectively decouple the analysis of their algorithms and data structures from the analysis of the memory buses and banks. Unlike specialized hardware customized for a specific data-plane algorithm, our system makes no assumption about the memory access patterns. We present a mathematical argument for our system's ability to provably provide bandwidth with high confidence and demonstrate its functionality and area overhead through a synthesizable design. We further show that, even though our scheme is general purpose to support new applications such as packet reassembly, it outperforms the state-of-the-art in specialized packet buffering architectures.

Improved global stability condition analysis for time-delay fast active queue management scalable transmission control protocol

“…The corresponding equilibrium point w 0 of equation (3) was uniquely determined as w 0 = cd + a. 4,5 Let R(t) 2 ½R 1 , R 2 , as a constant R c , we obtain as follows…”

“…FAST TCP 1,2 was proposed by Steven H Low et al of network laboratory of California Institute of Technology, and FAST TCP [1][2][3] was a new type of transmission control protocol proposed for the next generation of high-speed networks of high performance networks with high speed, long delay, and large capacity, and its global stability had attracted the attention of scholars. [4][5][6][7][8][9] Tan et al 9 and Koo et al 17 studied the global stability of the FAST system, but it simplified the case of time delay and time variation. In Koo et al 17 and Choi et al, 18 the Lyapunov-Razumikhin stability theorem and LMI method were used to obtain the global stability condition with lower conservativeness than that in Tan et al, 9 where time delay and time variation were not simplified.…”

“…Network parameters include propagation delay d and link bandwidth c. When these two sets of parameters do not match, the stability of the system will be affected, so it was necessary to establish the relationship between these two sets of parameters to guide the selection of protocol parameters and to find an effective way to solve the public problem. 4,5 In Agrawal and Sherwood 5 and Cui and Andrew, 6 a guidance scheme of selecting protocol parameters using static mapping table according to link bandwidth was proposed. Agrawal and Sherwood 5 pointed out on the basis of experimental simulation that in a single-link network, the protocol parameter a should be greater than 0.0075 times the link bandwidth.…”

“…4,5 In Agrawal and Sherwood 5 and Cui and Andrew, 6 a guidance scheme of selecting protocol parameters using static mapping table according to link bandwidth was proposed. Agrawal and Sherwood 5 pointed out on the basis of experimental simulation that in a single-link network, the protocol parameter a should be greater than 0.0075 times the link bandwidth. However, the above two schemes did not consider another network parameter propagation delay, which was an empirical value strategy, and it was difficult to ensure the stability of the system under different network environments.…”

Memory bandwidth reservation in the cloud to avoid information leakage in the memory controller

“…Virtually pipelined network memory [5] (VPNM) is a secure memory system that hides information leaks among threads that contend for memory banks in network processors. VPNM normalizes memory access time to trade off latency for deterministic, high bandwidth -good for network processing, but unsuitable for standard desktop/server workloads.…”