OpenFlow - what controller should I use - networking

I used the Floodlight for some openflow testing but I am not totally convinced by this and I saw that OpenDaylight is sustained by Linux Foundation and some of the major industry companies.
Can someone point some strong features that OpenDaylight have and FloodLight don't ? And if someone worked with OpenDaylight could point me to some good tutorials ?

have you checked out the Wiki? There is a lot of topics there. It is still getting sorted out and what not so you may need to use the search engine to find specifics more times then not.
We have some images that are pre-built OpenStack builds. The latest is from a couple of weeks ago for the upcoming OpenStack IceHouse release.
ODL and Icehouse Integration
All in all it is still rough around the edges and has a ton of development happening still.
There are some basic mininet examples using OpenFlow and the OVSDB plugin here:
OpenFlow and Mininet
OVSDB ODL plugin:
ODL and OVSDB Plugin with Mininet
For API related apps, Fred Hsu's blog while a bit dated now has some great examples:
OpenDaylight Api App usage
If you have any other specific interests the IRC channel is a good place to interact at irc.freenode.net #opendaylight or ping us in #opendaylight-ovsdb
Cheers,

This is the best one from the opensource available
http://sdnhub.org/releases/floodlight-plus-openflow13-support/
http://www.projectfloodlight.org/projects/
Hope this will help

Related

Writing an ML2 mechanism driver from scratch

I'm trying to implement an ML2 mechanism driver in order to be able to integrate a RINA stack implementation within OpenStack.
I'm struggling a bit to get started with the driver implementation, mainly because I'm quite a newbie with the openstack suite, and I am not able to find good documentation about implementing an ML2 from scratch. The only thing I found is this (old) book Openstack Networking Cookbook where on Chapter 10 it tackles the implementation of an ML2 mechanism driver from scratch. However, this book is based on openstack kilo (2015) and following the tutorial it just breaks the whole openstack installation.
So the main issue here is:
How should I get started? Should I just see how open source mechanism drivers are implemented and just try to guess how they work and start (slowly) to implement mine?
Is out there any good documentation or boilerplate code to implement an ML2 driver from scratch?
I only found these resources that seem relevant to the topic of writing ML2 drivers:
"Openstack Networking Cookbook" by Sriram Subramanian & Chandan Dutta Chowdhury (out of date).
"Writing your own OpenStack Neutron ML2 Driver for Cisco UCS" by Muhammad Afzal (Cisco Employee).
Neutron / ML2 on the OpenStack Wiki. Has links to a number of ML2 driver implementations / projects.
So to answer your questions:
How should I get started? Should I just see how open source mechanism drivers are implemented and just try to guess how they work and start (slowly) to implement mine?
That's probably the best approach. There are a few examples to look at.
You could also ask for suggestions and ask technical questions on the Neutron developer mailing list.
But you should expect to do most of the research for yourself; e.g. reading books, papers, specifications to get a deep understanding of networking, deep-diving the existing Neutron and ML2 codebase, etc.
Is out there any good documentation or boilerplate code to implement an ML2 driver from scratch?
Apart from what I have found, I doubt it.

How to setup the infrastructure with blade servers for OpenStack

We have 24 Huawei CH242 V3 blade servers and want to setup a private cloud with OpenStack, but we're very new to OpenStack and very lack of experiences about infrastructures. Could somebody kindly give us some useful information about the following question:
What kind of OS is more suitable for those blade servers? Is Linux like CentOS a good choice?
Is it OK(or encouraged) to directly use blade servers as OpenStack controller/compute/storage nodes? Or do we need to use one hypervisor to create many VMs and install OpenStack services on top of VMs?
What're the best practices or suggestions will you want to give beginners?
Maybe some questions are very silly but we're really stuck on the first step, thanks in advance for any information.
Below is my suggestions and there can be more good answers too
What kind of OS is more suitable for those blade servers? Is Linux like CentOS a good choice?
You can try any Linux flavours (OpenSUSE/CentOS/Ubuntu) mentioned in the openstack official site. I personally used Ubuntu for installing openstack.
There are openly available JuJu charms that works on Ubuntu for installing Openstack services. So it will be easy for you to edit the charms and deploy.
Is it OK(or encouraged) to directly use blade servers as OpenStack controller/compute/storage nodes? Or do we need to use one hypervisor to create many VMs and install OpenStack services on top of VMs?
I will prefer VM based installation from your list of choices. I personally suggest you to use containers to deploy your openstack services for better performance.
For compute service, you can go for bare metal installation, but it is upto you.
What're the best practices or suggestions will you want to give beginners?
a. Try installing the same topology/setup as mentioned in the openstack documentation
b. Use recommended databases and AMQP brokers
What kind of OS is more suitable for those blade servers? Is Linux like CentOS a good choice?
I use CentOS7.2, its very stable for openstack. and Ubuntu is also stable which is tried.
Is it OK(or encouraged) to directly use blade servers as OpenStack controller/compute/storage nodes? Or do we need to use one hypervisor to create many VMs and install OpenStack services on top of VMs?
Yes, I do like this, use bare machine as controller/compute/storage, performance good for me, I did not use container like docker.
What're the best practices or suggestions will you want to give beginners?
Because you are new to openstack, I recommend you begin with install openstack, see more logs when you install it. read official website docs is necessary. but you need to notice there are also some errors in the docs, and the configuration also is not optimized, that is just for experiment of private cloud.
If you are skilled at install openstack, then you can read the source code on github, try to contribute the code for it, from fix docs typo.

Why are the questions about Kaa basically not being answered?

I am not sure if this is a genuine question but I will try anyway.
The Kaa platform promises an open source IoT middle-ware apparently backed by a relatively large company. They suggest StackOverflow as the forum and place to ask about it, but for a few months basically no questions have been answered. Their older forum, also sounds kind of abandoned.
Does anyone have any idea why?
If not I would like to use this question/post to raise some awareness, discussion among enthusiasts, and hopefully hear the developers' voices.
I believe in an open source project it is always important to have some feeling about the state of the project, and the community behind it (which, as the company states on their website: should be vibrant) before committing some development/testing time. Kaa sounds as a great alternative for IoT projects. I would like to hear some of the opinions from other developers who are using Kaa about this issue.
Thanks!
pepgma, Mayur Patel,
Andrew Kokhanovskyi here, CTO at Kaa. First of all, please let me apologize for the lack of responses in public forums that you have experienced in the last few months. As you may know from our recent webinars, posts, etc., we are now working hard on the next version of the Kaa platform: codename "Banana Beach". It is because of the push to get the significantly improved version of Kaa to the early adopters that our engineering team has been having difficulties supporting all of the community requests lately.
Having said that, we truly appreciate your support of the Kaa open-source platform. As a token of gratitude, we would like to offer an invitation code: "AA-ADOPTER5" that you can use to register for the early access to Kaa BB here:
https://www.kaaproject.org/apply-for-early-access-to-kaa-1-0-banana-beach/
On our end, we will make sure that you will have a chance to be among the first ones to experience the new version of the Kaa platform.
--Andrew and the Kaa team

The best tool for deploying a production-ready OpenStack cluster

In your opinion, what is the best tool/solution to deploy a production-ready OpenStack cluster? It should be based on following criteria:
Stable and consistent across multiple deployments
Easy to maintain and upgrade
If possible, intuitive and easy to customize for different scenarios
This really depends on your knowledge of Openstack, For Senior Openstack experts I recommend Ansible, You will write your own playbooks, with this approach you have full visibility on what you are doing.
For Less experienced openstack administrators:
Packstack
Mirantis
SuseCloud
VMware Integrated Openstack
...

Voting system/engine for customers?

I'm talking about some web thing like http://uservoice.com/
Can you suggest any other similar service, web-site or may be (even better) a ready engine for deployment on own server?
Actually, the question more about systems, which can be installed on your own server.
UseResponse, commercial (full sourcecode available on purchase), launching December 2011, with live demo available on USWebStyle website. Fully customizable (design, functionality).
Types of feedback (idea, problem, question, thanks), vote types (positive and/or negative) are adjustable.
Installable on any PHP 5.2/5.3 hosting environment.
TenderApp seems to have a lot of the same features, but it's also SaaS.
KBPublisher can be installed on your server.
By the way, most of these SaaS systems like UserVoice will let you forward your own domain/subdomain to their service and apply custom branding, so the experience to your end user is very similar to being on your actual site.

Resources