An adaptive hash-based multilayer scheduler for L7-filter on a highly threaded hierarchical multi-core server

Abstract

Ubiquitous multi-core-based web servers and edge routers are increasingly popular in deploying computationally intensive Deep Packet Inspection (DPI) programs. Previous work has shown the benefits of connection locality-based scheduling on multi-core servers to improve L7-filter performance. However, we show that highly threaded hierarchical multi-core processors, such as the Sun Niagara 2 processor, accumulate imbalanced workload at each resource layer. This workload imbalance potentially offsets the benefits from connection locality. In addition, connection-locality-based load balance fails to work when network traffic is unevenly distributed. In this paper, we propose an adaptive hash-based multilayer scheduler for a highly threaded hierarchical Sun Niagara 2 server. Our scheduler maintains connection locality and adaptively adjusts the scheduling to balance the real time workload. The original Highest Random Weight (HRW) hash guarantees the connection locality but only balances the workload over the number of different connections. We enhance the original single layer HRW into a hierarchical "hash tree" scheduler to balance the connection workload in accordance with the hierarchical processor architecture. We then optimize our multilayer scheduler to adaptively adjust scheduling decisions based on service time at each level, further improving the system load balance. Our scheduler is shown to increase the system throughput by 59.2% compared to the previously proposed connection locality optimization.