Ask Your Question

Relationship between OPNFV architecture and Pharos compliant lab structure. [closed]

asked 2015-09-02 05:44:33 -0700

Deivarayan gravatar image

updated 2015-09-02 06:07:21 -0700


I am quite new to OPNFV and am trying to understand the OPNFV hardware architecture and its requirement.

OPNFV architecture requires 3 nodes (Compute, storage and network node). Pharos compliant lab structure requires 5 nodes (1 Jump server, 3 Controller nodes and 2/n compute nodes) according to Pharos hardware spec.

Comparing these 2 architecture i have the following queries. It will be very helpful if someone could help me in understanding these.

  1. Requirement of 3 controller nodes ? Does one of the controller nodes in pharos architecture act as a storage node and the other controller nodes has different functions?

  2. In which component (JumpServer/Control nodes/Compute nodes) in pharos architecture, does the VNF run (With respect to processing and storage)?

  3. From the hardware architecture guide of pharos compliant lab structure, it looks we require disk space of 2*1 TB + 1 * 100 GB SSD and 32GB RAM. Is it a minimum requirement or can we still have a lower disk space and memory for basic lab setup ?

  4. Please advice on the minimum disk space and RAM requirement for each components of Pharos (Jump Server, Control nodes and compute nodes).

  5. Does management functions like VNF manager and VIM runs (Processing wise) on Jump server?

Thanks in advance!!!!

edit retag flag offensive reopen merge delete

Closed for the following reason the question is answered, right answer was accepted by Deivarayan
close date 2015-09-03 01:50:33


you're welcome - can you mark the question as answered? thx!

lmcdasm ( 2015-09-02 14:33:37 -0700 )edit

1 answer

Sort by ยป oldest newest most voted

answered 2015-09-02 11:30:59 -0700

lmcdasm gravatar image


i have tried to answer your queries below - however, i think your issue is one of context. A Pharos lab when being put to use for CI pipeline requires more nodes than the basic Reference Architecture - which is referenced here (i think you meant Openstack when you said Compute,Storage and Network - and actually its "Compute, Control" in terms of Openstack nodes.. can you link the architecutre you are referring to (since "Compute, Storage and Network node" is not quite right).

At any rate, i have tried to answer your questions here:

Answer 1 - If you want to run your openstack deployment in an HA (High Available) Manner, then you need to have three controller nodes in operation. The three controllers act as openstack controllers. We dont implement a storage node nor network node in OPNFV deployment archecture currently - but the framework in the architecture can be expanded easiely (the ip subnets are in place) - if you wanted to move storage off the compute/controllers (currently CEPH, and LVM are available in FUEL for example).. you can move the neutron components off a controller if you wanted to as well - but this is currently not the architecture..

here is the architecture reference link for OPNFV - maybe you can refine your question a bit to match the arch (or provide the link to what you are looking at?). Here is the link to the Pharos state (with links to the reference architecture)

Answer 2 - im not sure what you mean - but if you are asking where the VM (that is the VNF) runs, well its in the deployed openstack environment on one of the pharos labs - so you expect that a pharos lab deploys an installer and a openstack environment and the VNF is then deployed in there - after all your VMs run inside the Openstack environment that is deployed.

Answer 3 - Can you show me the link to what you are looking at? I would need to see in what context / what node you are talking about.

Answer 4 - refer to the installer that you are choosing release notes, they have the requirements per installer listed there.

Answer 5 - no - there is no VNF manager (since that would be managing the VMs inside openstack), FUEL provides management of the metal resources (on which either a control or a compute node runs). A VNF manager would be a different node in the network (like Tacker or something).

Cheers, D

edit flag offensive delete link more

Question Tools

1 follower


Asked: 2015-09-02 05:44:33 -0700

Seen: 142 times

Last updated: Sep 02 '15