Ask Your Question
0

Neutron ML2 questions

asked 2013-12-26 20:36:13 -0500

chen-li gravatar image

updated 2014-01-22 15:12:26 -0500

Evgeny gravatar image

Hi list,

I just read the presentation slides showcased in the Openstack Hongkong summit, on "Openstack Neutron Modular Layer 2 Plugin Deep Dive". http://fr.slideshare.net/mestery/modular-layer-2-in-openstack-neutron

I have some questions, hope someone can help me.

Q1. P7 : there is a new feature "top of Rack switch control", what is a rack switch means ? Why ML2 can do top control of it ?

Q2. P11: It said "Allows multiple mechanism drivers to access same network simultaneously", do we have some examples to explain this ?

Q3. P13, what steps should I do to achieve this kind of network ? Any guide we have ?

Looking forward to your reply.

Thanks.

-chen

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted
1

answered 2013-12-27 04:17:42 -0500

nyechiel gravatar image

Q1. Top of a rack switch (ToR) is a physical switch located inside the rack and provides connectivity to the hosts in the rack. It is usually directly connected via a couple of 1G or 10G Ethernet links to the hosts NICs. You can find ToR switches from various network vendors such as Cisco, Brocade, Juniper, Arista, etc.

ToR is probably the most common implementation in modern data-centers, but it's worth mentioning that there are also the concepts of middle of rack (MoR) and end of rack (EoR) designs. You can find more information on the Internet, but this is a good starting point: http://www.excitingip.com/2802/data-center-network-top-of-rack-tor-vs-end-of-row-eor-design/

Utilizing software defined networking architecture, the idea is to control the ToR (physical) switch and program its forwarding entries the same way you can control vSwitches using protocols like OpenFlow and OVSDB. This is mostly beneficial to ease the provisioning time of the network and provide access to bare-metal (physical) appliances to the logical network. There are some vendors working in this direction, and they can offer their solution as a driver for ML2.

Q2. With the original implementation of Neutron, you can also had a one (monolithic) plug-in to choose from. This limited your choice when it comes to the overall network solution. For example, if you had an environment with KVM hosts running Open vSwitches, and then physical network devices from Cisco and Juniper - you did not have the option to control them all using Neutron. Practically, you had to choose one plug-in (for example the OVS), and manually provision the rest of the network (Cisco and Juniper devices in this case). The idea behind ML2 is let you choose (and mix) your network solution, and integrate with one ore more vendors using the MechanismDrivers.

Q3. It is really hard to answer. I guess you need to refer to the different vendors offerings and guides.

Hope it helps, Nir

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower

Stats

Asked: 2013-12-26 20:36:13 -0500

Seen: 741 times

Last updated: Dec 27 '13