This paper presents vCAT, a novel design for dynamic shared cache management on multicore virtualization platforms based on Intel's Cache Allocation Technology (CAT). Our design achieves strong isolation at both task and VM levels through cache partition virtualization, which works in a similar way as memory virtualization, but has challenges that are unique to cache and CAT. To demonstrate the feasibility and benefits of our design, we provide a prototype implementation of vCAT, and we present an extensive set of microbenchmarks and performance evaluation results on the PARSEC benchmarks and synthetic workloads, for both static and dynamic allocations. The evaluation results show that (i) vCAT can be implemented with minimal overhead, (ii) it can be used to mitigate shared cache interference, which could have caused task WCET increased by up to 7.2 x, (iii) static management in vCAT can increase system utilization by up to 7 x compared to a system without cache management; and (iv) dynamic management substantially outperforms static management in terms of schedulable utilization (increase by up to 3 x in our multi-mode example use case). Abstract-This paper presents vCAT, a novel design for dynamic shared cache management on multicore virtualization platforms based on Intel's Cache Allocation Technology (CAT). Our design achieves strong isolation at both task and VM levels through cache partition virtualization, which works in a similar way as memory virtualization, but has challenges that are unique to cache and CAT. To demonstrate the feasibility and benefits of our design, we provide a prototype implementation of vCAT, and we present an extensive set of microbenchmarks and performance evaluation results on the PARSEC benchmarks and synthetic workloads, for both static and dynamic allocations. The evaluation results show that (i) vCAT can be implemented with minimal overhead, (ii) it can be used to mitigate shared cache interference, which could have caused task WCET increased by up to 7.2×, (iii) static management in vCAT can increase system utilization by up to 7× compared to a system without cache management; and (iv) dynamic management substantially outperforms static management in terms of schedulable utilization (increase by up to 3× in our multi-mode example use case). Disciplines Computer Engineering | Computer Sciences

Fine-grained network traffic monitoring is important for efficient network management in software-defined networking (SDN). The current SDN architecture, i.e., OpenFlow, relies on counters in the flow entries of forwarding tables for such monitoring tasks. This is not efficient nor flexible since the packet-header fields that users aim for monitoring are not always the same or overlap with those in OpenFlow match fields, which is designed for forwarding as a higher priority. This inflexibility may result in unnecessary flow entries added to switches for monitoring and controller-switch monitoring-based communication overhead, which may cause the communication channel to become a bottleneck, especially when the network includes a large number of switches. We propose SDN-Mon, a SDN-based monitoring framework that decouples monitoring from existing forwarding tables, and allows more fine-grained and flexible monitoring to serve a variety of network-management applications. SDN-Mon allows the controller to define the arbitrary sets of monitoring match fields based on the requirements of controller applications to flexibly monitor traffic. In SDN-Mon, some monitoring processes are selectively delegated to SDN switches to leverage the computing processor of the switch and avoid an unnecessary overhead in the controller-switch communication for monitoring. We implemented SDN-Mon and evaluated its performance on Lagopus switch, a high-performance software switch.

No abstract

No abstract