18. 18www.rackspace.com
PNF, VNF, NFV, SR-IOV, DPDK
Physical Network Function
Virtual Network Function
Network Function Virtualization
Single Root I/O Virtualization
Data Plane Development Kit
19. 19www.rackspace.com
NFV
(Programmatically define and execute the services that run on
top of those SDN networks)
|
VNF
(The actual building blocks your NFV infrastructure
manipulates)
|
SDN
(Programmatically define your networks)
|
OpenStack
25. 25www.rackspace.com
Supporting Telecom and NFV with the Open Source
Ecosystem
John Zannos and Tobias Ford presented this at the Paris OpenStack Summit
https://www.openstack.org/summit/openstack-paris-summit-2014/session-
videos/presentation/supporting-telecom-and-nfv-with-the-open-source-ecosystem
26. 26www.rackspace.com
NFV and Telco Solutions on OpenStack
Maarten Ectors and crew presented this at the Paris OpenStack Summit
https://www.openstack.org/summit/openstack-paris-summit-2014/session-
videos/presentation/nfv-and-telco-solutions-on-openstack-canonical
28. 28www.rackspace.com
NFV, VNF, and Deterministic Architecture: What Does It
Mean?
Nick Chase
https://www.mirantis.com/openstack-portal/external-news/nfv-vnf-and-deterministic-architecture-
what-does-it-mean/
30. 30www.rackspace.com
For those of you currently working for a telco, are there
any use cases you are using OpenStack for today that I did
not talk about?
32. Thank You!
I’m jthorne on Freenode.
I’m jameswthorne on Twitter.
I blog at thornelabs.net.
Editor's Notes
I'm James Thorne.
Cloud Solutions Architect at Rackspace where I've worked for the past 2 years.
Today I will be talking about telco OpenStack uses cases. I am going to discuss Rackspace's experience working with telcos who want OpenStack, delve into why they are interested in it, what they want out of it, and the challenges faced.
Now, I’m by no means new to OpenStack, but I am new to the world of what telcos want out of OpenStack. I’ve spent the last month talking to people at Rackspace, watching previous OpenStack Summit videos, and reading blog posts to figure out how everything fits together.
What I’ve discovered is that what telcos want out of OpenStack is just as complicated as OpenStack on its own. This presentation is very much a new comers view of the telco OpenStack ecosystem. So, if I get something blatantly wrong, feel free to correct me, but please correct me at the end of the talk during the Q&A session.
Before we dig into the content, I want to ask you all several questions. Don't worry, all you have to do is raise your hand.
OpenStack just recently turned 5 years old. 5 years young is probably a better way to say it.
OpenStack's growth has been nothing short of phenomenal in those 5 years, and its growth is only increasing.
You sometimes hear about hockey-stick-like-growth in the stock market, well that sort of growth is happening to OpenStack. Especially in the enterprise space.
Enterprises have really started to pay attention to OpenStack over the past year.
When I joined Rackspace, I was mostly working with small to mid-size companies. Now, more and more, I am talking to enterprise companies. Of course, included in those enterprise companies are telcos.
I believe a large part of this adoption is because of three specific things.
Reliability, stability, and upgradeability.
Beginning with OpenStack Icehouse, the OpenStack Community really began to focus on these three things. Code was introduced to allow easier and more reliable upgrades. And since then, these three areas are constantly improving, and it’s really start to pay off.
Reliability and stability are a given for any platform, but what really maters to enterprises is the ability to easily upgrade.
We all know OpenStack has a 6 month release cycle. And we all know Enterprises do not upgrade every 6 months. Being able to upgrade an OpenStack environment with little to no downtime will spur even more enterprise adoption.
So, with all that said, before I dive into why telcos are interested in OpenStack, what are some of the characteristics of a telco?
Telcos want to control the entire stack - from the hardware to the software.
Even though they want to control the hardware and the software, they also want someone to lean on when something goes wrong. They want to be able to pick up the phone. They want one-throat-to-choke. They want professional service engagements when they need them.
Telcos have traditionally relied on a single vendor to help them build the entire stack for a particular product. They will of course work with more than one vendor across their enterprise, but a particular infrastructure, for example the infrastructure for SMS, may all come from one vendor.
A single vendor will provide an entire stack that consists of very expensive, specialized devices that have different uses. Without all of them, the entire stack doesn't work. This stack is used for some number of years and may not change. It is hard to innovate here. It is difficult to move quickly.
Because of the hardware and software lock-in, telcos can only iterate every 5 - 10 years.
So, finally, why do telcos want OpenStack? There are two big reasons.
First, they want it for the same reason everyone else does. To become a service provider for their internal or external customers.
Telcos, and enterprises alike, are tired of it taking days or weeks to get resources allocated, whether that's compute, storage, or networking. Because it has traditionally taken so long to get resources, developers and whomever else faced with short deadlines are forced to take the path of least resistance and use public cloud resources. There’s obviously many problems with this. Security concerns and leaking intellectual property are just two examples.
Telco and enterprise IT departments need to offer cloud platforms and resources inside their own walls and firewalls. OpenStack allows them to do that.
Not only does OpenStack allow IT departments to offer cloud resources inside their own walls, those managing the cloud platform suddenly have a standard platform for almost everything. Whether that’s for internal customers hosting development environments or internal applications, or for external customers accessing public facing websites or applications.
So, telcos want this, but why do they need this? Also just like everyone else, so they can iterate faster. In today’s world if you can’t iterate quickly and get products to market, you’re going to be left behind. A great example is SMS. SMS revenues are down drastically. Why? Because of products like Apple Messages and WhatsApp. Those products are a boon to consumers because they work globally, are more featureful, and don’t really cost anything. Plus, they’re cool, but that’s a whole other topic of discussion.
If a telco’s SMS infrastructure is locked into a single vendor’s hardware and software, it is exceptionally difficult to add features. Telcos want a standard platform where they can take their physical network functions and turn them into virtual network functions. OpenStack can and will provide that.
And this brings us to the second use case that typically doesn’t apply to the rest of us. There are specific functions telcos have to implement that none of us have to worry about. Now we enter a whole new world of acronym soup.
PNF, VNF, NFV, SR-IOV, and DPDK. And that isn’t even all of the acronyms you’ll come across, but they are some of the main ones.
I’m going to briefly describe each of these things. Digging into each of these would require another 45 minute presentation.
Physical Network Function is the current and legacy way of doing things. Telcos purchase a specialized device from a particular vendor. Something like a hardware load balancer for SIP traffic.
Virtual Network Function is virtualizing the specific functions the specialized hardware devices would do in software. Instead of purchasing a hardware load balancer for SIP traffic from a particular vendor, a telco could purchase commodity servers and load balancers and implement the load balancing in software.
Network Function Virtualization is the network architecture concept of virtualizing entire network functions by using the Virtual Network Functions as building blocks to create some sort of communication service. The building blocks would be chained together to create something called service chaining, which is another term you will come across.
The next two terms are less concepts and more specific things that make all this possible.
Single Root I/O Virtualization provides virtual machines direct access to a physical ethernet port. For example, the Intel X560 NIC provides 32 channels. Each of these are raw channels that any protocol stack could be run on top of. If you want to dig deeper, Scott Lowe wrote a fantastic blog post on SR-IOV back in 2009. Some of these technologies are hardly new. I’ll provide the link to that post in the references section.
Data Plane Development Kit is a set of data plane libraries and network interface controller drivers for fast packet processing. So, it’s essentially a programming framework. It helps telcos extract as much performance as possible out of the commodity gear.
So how do all these work together?
For me to understand something, I need to be able to visualize it in my head. Most things in IT can be stacked on top of each other.
This is how I believe these things fit together.
In addition, I’ve seen telco OpenStack architectures that remove SDN completely with the NFV and VNF concepts running on their own.
All of these things that most of don’t have to worry about will allow telcos to implement everything they currently do on top of a standard platform.
So, now that we have discussed why telcos want OpenStack, what have been some of the challenges for them adopting it? Why has it taken this long for them to get on-board?
Once again, reliability, stability, and upgradeability. Every time we make a phone call or send a traditional text message, we expect it to just work.
Telcos are not going to rush into something that isn’t reliable, stable, and upgradeable. Telcos need to be sure what they are using is carrier grade.
As mentioned earlier, OpenStack Icehouse really started to focus on these changes and it’s only getting better with each release.
Telcos have a lot of intellectual property. It’s a tough industry and they will do anything to keep that competitive edge. Unfortunately, that typically means there are little to no contributions to open source software.
Both telcos and typical enterprises are slowly warming up to open source, and more and more of those companies are contributing.
I think as more of the mainstream telcos begin to contribute to open source, everyone else will jump on the bandwagon that much faster.
Telcos are used to iterating every 5 years, not every 6 months. How can something that moves so quickly be reliable?
A lot of companies are running their internal and external production applications on OpenStack. I work and talk with them everyday. Of course there are always problems, but the platform is currently stable enough for those companies to run their business on it.
As the OpenStack platform continues to mature and more companies go public about their infrastructure, telcos and enterprises alike will begin to take notice and make the transition.
Content wise, that is all I have for you today. I wanted to quickly run through some of my references which includes further material for you to watch or read if you want to dive deeper into the world of telcos and OpenStack.
Before I open the floor for Q&A, I'd like to ask the audience one more question. This time I'm looking for more than just hand raising.
Great! Thank you for your input. I'll now open the floor for Q&A.
If for some reason I cannot answer the question, I welcome anyone in the audience who can answer it to speak up.