Please note: All times listed below are in Central Time Zone
Openstack based private clouds offer an ideal environment for operating scientific workloads. Abilities like flexible resource allocation, self service and administrative control are exactly what is needed to offer agility and speed up research. However, in practice Openstack is rarely used in such environments. When it comes to scientific applications, there is hardly any documentation on techniques and architecture for Openstack based cloud.
In this talk, we provide the missing guide.
- Workload Classification: CPU, Memory or IO bound.
- Managing CPU w/ Openstack, libvirt and KVM: Advanced CPU features, NUMA topology, Cache coherency
- Managing memory: over-provisioning, avoiding pitfalls w/ page sharing, NUMA architecture
- Network I/O: Tunneling overhead and Security Groups. Employing SR-IOV
- Managing users and access to resources for Scientific Cloud
- Opportunistic scaling of computational applications
- Streamlining research with Murano apps versioning
* Tweaking OpenStack to extract baremetal performance for Scientific Computational Workloads
* Planning OpenStack deployment to match requirements of workloads with underlying physical resources
* Architecture of Scientific Computational Cloud