Throughput-delay analysis of interrupt-driven kernels with DMA enabled and disabled in high-speed networks

Abstract

Interrupt processing can be a major bottleneck in the end-to-end performance of high-speed networks. The performance of Gigabit network end hosts or servers can be severely degraded due to interrupt overhead caused by heavy incoming traffic. Under heavy network traffic, the system performance will be negatively affected due to interrupt overhead caused by the incoming traffic. In particular, excessive latency and significant degradation in system throughput can be experienced. In this paper, we present a throughput-delay analysis of such behavior. We develop analytical models based on queueing theory and Markov processes. In our analysis, we consider and model three systems: ideal, PIO, and DMA. In ideal system, the interrupt overhead is ignored. In PIO, DMA is disabled and copying of incoming packets is performed by the CPU. In DMA, copying of incoming packet is performed by DMA engines. For high-speed network hosts, both PIO and DMA can be desirable configuration options. The analysis yields insight into understanding and predicting the impact of system and network choices on the performance of interrupt-driven systems when subjected to light and heavy network loads. Simulations and reported experimental results show that our analytical models are valid and give a good approximation.