Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

RDO - looking for architecture recommendation

We are looking to deploy a highly available RDO Havana solution within our existing development, integration, and test environment. We are looking to design a multi-tenant solution where each tenant can configure and deploy their own private networks and use our provider networks in the data center for internet access as well as inter-tenant communication. We are thinking of using gre tunnels for tenant private network separation.

We are looking for a two nic solution in the design to handle the openstack (management, data, external, and API networks), plus any others we may need to integrate into our current environment.

We are looking for an RDO recommended network architecture which maps each of the openstack networks to physical nics. We want to be solid on the underlying nw architecture before building the solution. In the past with Folsum and Grizzly, we starting deploying a solution and later got burned with unforseen openstack requirements that required band aids to the nw solution. we would also like a recommendation on which services map to each server (assume 5 to 14 blades total). we are concerned with examples which use single network nodes due to bandwidth and availability limitations, assume 100% of our tenant VMs will require internet access. We have seen examples where each compute node runs Neutron L3 services, is that feasible?

Our server hardware consists of Blade servers with two 10Gbit nics and two 8Gb SAN HBAs. The blade chassis contains dual 10Gbit switches which are not currently stacked. We are limited in this respect because not all of the blades in the chassis are being allocated to RDO. the blades are 2cpu/6core with 72Gb memory and no disk, we use boot from SAN for the RHEL OS.