A Study On Optimization Of Network Resource Consumption For Middlebox

Middleboxes are ubiquitous today in network. They perform a variety of functions, including firewalling, intrusion detection system, Load Balancing, BRAS and WAN optimization. These functions are usually deployed on exclusive hardware, which needs manual maintenance and is difficult to manage or update due to their heterogeneity, and therefore generates great cost for network operators. To address this issue, the conception of Network Function Virtualization (NFV) is proposed. NFV improves the flexibility in function expansion and reduces Middlebox cost., Given the promise of cloud computing, NFV is on its way to be deployed on cloud platforms. However, the high workloads and the multipath feature in cloud, making schedule different resources under NFV scenario becomes a difficult problem, and CPU to be one of the bottleneck, especially when overwhelming network flows exist. Under the NFV scenario, the requirement of reducing resource consumption, especially the CPU resource, addresses a lot of attention.Existing efforts mainly focus on the optimization of flow scheduling and Middlebox re-deployment. These efforts cannot adjust their deployment dynamically according to network status, and therefore fail to meet the requirements of flexible deployment in enterprise network. Based on our previous work, we propose a new solution that is, to adjust a Middlebox function dynamically by monitoring the real-time change in network flows in order to reduce resource consumption in cloud. On a Click platform we deploy and implement three Middleboxes using our solution framework:BRAS, Load Balancing and IPS. Experimental results show that our solution can reduce CPU utilization by 21.9% and improve network throughput by up to 23.5% compared with without our solution, therefore greatly alleviate the CPU overloading and network congestion problem.