LCE-EU 2015

With simultaneous threads competing for processor shared resources like
cache, it could result in cache contention leading to negative impact on
key Quality of Service guarantees. In this talk, Vikas will discuss
using cache monitoring and allocation support in Linux kernel based on
perf and cgroup, which provides a lightweight, powerful tool to address
such cache QOS issues by providing reasonably accurate ways to monitor
and allocate cache. Vikas will discuss challenges faced to using the
feature in cloud, containers context and clients. The talk will discuss
enterprise benchmark data focussing on cache contention issues, the
measured improvements using the feature, the challenges involved in
Kernel to overcome limited hardware resources,allocating cache resource
per thread,per socket,issue of segregating instruction and data in
allocations and issues in context switching.