We believe not all environments behaving the same. As private cloud owner, Having an infrastructure that is flexible enough to be configured in different models is likely the right direction to enable and empower your own applications, at the end, what we want enable is all about how application perform, and how our business perform based on all server side optimizations we are doing.
In this talk, we will explore multiple options within modern operating system which potentially give us opportunities to fine tune system performance in addition to generic cloud interface based on VM characteristics, we will also share the way we are doing in Cisco WebEx cloud incubation to achieve the goal of VM performance optimization, and also share the result and recommendation with the people who care more about applications than fancy cloud features.
In this talk we will go over below items as overview for VM performance optimization
- Most of Openstack cloud using KVM as fist tier hypervisor, guest VM in such case run as a Linux process on the host.
- Virtual CPUs (vCPUs) are implemented as normal threads, handled by the Linux scheduler.
- Guests inherit features such as NUMA and Huge Pages from the kernel.
- Disk and network I/O settings in the host have a significant performance impact.
- Network traffic typically travels through a software-based bridge.
With all that in mind, mantain a meanful selected scope of tuning parameters based on VM characteristics is something pravite cloud owner could think about, a small demon session running within compute host will offer us the possibility to tune the VM performance automatically, also nova extended framework will enable an end2end tuning flow in any given cloud ecosystem.