Summary:

The fear of cloud lock-in still dogs broader cloud adoption.


Transcription details:
Date:
20-Jun-2013
Input sound file:
1002.Day 2 Batch 1

Transcription results:
Session Name: Compute Everywhere: The Architects Behind the Change

Joe Weinman
Jordan Novet
Chandra Pandey
Lane Patterson

Joe Weinman 00:00
We’re kicking off our day with something I think is pretty exotic, which is we’ve been talking about cloud now– this is the sixth Structure ever, and the first Structure in 2008 was really kind of the defining start I would say of the cloud era. But really people have been talking about cloud and they’ve really been offering something that looks a lot more like mere utility computing. So the notion of taking a disbursed fabric of facilities and creating a single logical whole out of it, has a lot of implications both in terms of resource allocation and in terms of sophisticated network technologies.
Joe Weinman 00:39
And so we’re going to kick off today with a bang, with some of the top experts in the world on this subject. So without further ado, let me welcome Jordan Novet a staff writer with GigaOM and his two amazing guest.

[applause]

[music]
Jordan Novet 01:03
Good morning everybody. I’m pretty excited, you can see I was really eager to jump out on the stage, because we have an SDN use case. We’ve been talking about SDN with Bob Muglia from Juniper yesterday. We’ve been talking about how VMWare spent $1.26 billion to get Niciar, and how that’s going to help Gelsinger and Co. get ahead in the hybrid cloud efforts. And now what really is puzzling to me though is how long is it going to take for enterprises to get on board with SDN, and whether services providers are going are also going to make this part of their schtick.
Jordan Novet 01:51
So here today, we have two guys who are representing companies that are working on how to do SDN, not inside of the data center, but among them – on the Metro and then on a wider scale. So, is it going to take a long time for this to happen Chandra?
Chandra Pandey 02:14
If you look into the SDN, what the SDN has done it– a lot of IT people have looked into it and said, Stanford and a lot of companies– people are working on creating the controllers, and we know that it’s going to take a time, the approach is the right one. Because that approach kind of creates the functionality, where you get the scalability at the cloud level. You are not tied up with just a basic platform in there. And we do see that a lot of good things are going to come out and we will have the deployment, some deployments are already in the seller state, but there are more going to come over the time. But just in this current form, no. There will be a lot of evolution and refinement happening in there.
Chandra Pandey 03:07
As far as like BPI is concerned, we sort of kind of understood what the service providers are looking for. Kind of like the cloud based services delivery requires a different set of the network altogether. And those networks do not have to be just kind of the way it was done before. Cloud based services are hosted in the cloud and those you should be able to access no matter where you are in the world, and no matter what kind of device you have in there. And that requires the network to have an intelligence, it has to know where the service is coming from, and you have to like think different then what approach was taken before.
Chandra Pandey 03:53
So what the BTI did, just knowing that industry is going to move in that direction, no matter what. May not be current form, but there is a form which is going form. So we came with the concept, and we’re pretty lucky to be working with some of the carriers who always look forward and they are trying to solve the problems for the future, not just like the problems of the past. So that gave us like a visibility of what needed to be done, and the intelligent cloud connect platform be made – it’s an open platform. But at the same time, we also knew that not all the layers has to be at every places. There are locations in the network which just cared about having the expressway. Then there are locations that cares about having the highway and then the service-way. And service-way is the way it does not have to be sitting into the network platform, it can be sitting into the cloud anywhere around the world and creating a one fabric which enables you to deal with the services from anywhere. And I’m pretty sure that Lane is going to talk about– we were just talking backstage that he wants to create this marketplace fabric, where the services can be delivered from anywhere and they want to have a zero maintenance window. And that is what’s required. There is no maintenance windows. Yesterday some of the CIOs mentioned that right now the way they deal with the services, the way they have their global workforce, there’s a zero maintenance window. And how you enable that? That’s where like your SDN network virtualizations, as is well like the services delivery is going to take a place. And it requires a platform where you are able to like go and see at the different layers, and at the same time have an intelligence where you are going to deliver the service from. That is what’s going to matter, I think.
Jordan Novet 05:56
And just to be clear, you keep using the word platform, I think people might be confused. What kind of a platform are we talking about? I know there’s some gear, I know there’s some software, so just explain that a little bit.
Chandra Pandey 06:05
So what we are talking about is like definitely there is a network gear involvement there. But the gear is not just like tied up everything has to be there and that is when we talk about the service switch. We’re talking about having the APIs to go and control the platform.
Jordan Novet 06:25
That’s not exactly yet what Equinix is doing. What exactly is Equinix doing?
Lane Patterson 06:30
So the point Equinix is at right now, certainly our vision is dynamic interconnection infrastructure. And we’ve started by choosing BTI for really kind of a best in bridge for the optical layer, but with a strong packet roadmap on top of that. With kind of an open flow 1.3 approach and restful API and so on. And so, all of the platform choices that Equinix is making right now are really being made toward automating for structure. Not just for our own internal needs, but we really represent a place where– what Equinix really is to the market– let me go back all the way up to the use case level, because I think people want to know why we want to do this. What kind of value are we creating for our customers? What kind of business problem are we solving?
Lane Patterson 07:21
The reality is that the cloud has been validated, we’re going to be in a hybrid cloud environment for the next 10-15 years, and apps are distributing. So people want access to their apps to the nearest local data center. So if you’re in Europe, you want to be able to access your apps at a Frankfurt data center or a London data center. If you’re in North America, you want your apps to be east coast, west coast. And if you’re in Asia, you want your apps to be Singapore, Hong Kong, Tokyo. Wherever, approximate to where you are, so you get the best possible performance.
Jordan Novet 08:01
But you’re not doing this across the world yet?
Lane Patterson 08:04
Right.
Jordan Novet 08:05
You’re just starting it in one little corner of the world.
Lane Patterson 08:07
So we have three or four cloud hubs now. The name of the game for Equinix centers is we’re kind of the ultimate open infrastructure platform. All right, we’re kind of the platform where the network crew up or the internet kind of grew up to scale to each other. Then Web.10 and content and CDNs kind of grew up around this Equinix open model, and now enterprises are and I guess for the last few years cloud providers have been building in with us. We’re now seeing enterprises anchor themselves at network neutral facilities. And the reason why is for two things, they want to get their WAN performance hubs, they want to build out their WANs in open architectures where they can leverage best in breed providers and best in breed services. And they want direct access to the cloud. So CIO is a portfolio manager and today maybe only the bottom top 5% or 10% of their portfolios for a typical large company can be put up in the public cloud. They want to trust the cloud for more. One of the things holding us back is using an unpredictable kind of unguaranteed internet path. And what Equinix is all about is using its Ethernet fabric with kind of SDN automation on top of it for any customer that comes in to have access to a common fabric where they can automatically signal themselves a path to any cloud provider, any network provider. So if an enterprise comes in and wants 10 megs of SIP Trunking or wants access to ADWOS or access to Azure or access to SalesForce, any of these large clouds. Maybe they want a DIOS internet access product, maybe they want to get access to a WAN, a MPLS or a carrier ethernet product. All of those things could be available through a single port dynamic API driven model.
Jordan Novet 10:05
So Chandra, how easy is it for me to use this if I’m an enterprise, I’ve got a cage in Equinix, how easy is it for me to dial in with this kind of intelligent cloud connect thing? I want a certain amount of bandwidth to get to AWUS to enable that hybrid cloud access and not have to deal with how slow the public internet is going to be.
Chandra Pandey 10:25
So as Lane pointed out, he has already created this marketplace where the enterprise is coming as well as the cloud providers coming. Actually I had the opportunity to talk with him sometime back and he had this concept of 6 & 10. And in the 6 & 10 what he’s doing is bring the 6 – the biggest cloud provider in his marketplace and then the 10 geographical market where he can go and address those. And that’s what basically is like for having the open platform, as Lane pointed out, the restful API and the SDN fabric framework, it’s pretty easy. People can have their own portal and then they can go in and just press the button. That work has to be done right now on the head side, as well as some work has to be done on the industry side as well, but we aren’t very far away. And one thing I want to say, this is something like being a start-up company, per se. As a start-up we don’t have to wait for all the standards to get finalized. Some of the things we have to do since these are software where there is a format. We take the step and we do understand that sometimes the standard may not be exactly the same. But that’s going to be just like format upgraders, software upgrade. And those are the some things you always have to do. Same thing I’m talking about, it’s almost 11 years back, carrier ethernet that was just a talk. And as a company called Internet Photonics we went ahead and defined what the carrier ethernet is going to be. And it took the industry seven or eight years to get to the same point. But it was done in the form where we can go and upgrade it and make it a standard. Same approach we took, right now if you look into like there’s a lot of extensions that are not specified for the MPLS or even like the optical wavelengths, sub-wavelengths, circuits, and all that. I got my team one day in my office and said, “Guys, look we are going to do this.” Just go do it and the standard will take some time and we will go and confront today’s standard, but it’s a software update. We have to go and do that. So a lot a time that’s what we bring on the table. When we talked with Lane and he shared his vision, And this is what we want to do, and we bought into that – we are on the same page. We will go and make it happen. That’s what we’ll do.
Jordan Novet 13:02
Okay, so it sounds like this could have some type of business advantage in terms of SLAs, is that right Lane?
Lane Patterson 13:09
Yes, absolutely. I think the whole vision is for the cloud to scale to the mission critical applications and predictable performance for a large enterprise, to really mature beyond the SND segment. And the web internet, the kind of web hosting SAS segment, you have to be able to guarantee performance. And cloud providers are getting much better at guaranteeing performance levels, measuring and monitoring fairness among VMs, and virtual access to storage. On those two legs of the stool, but network is the third leg of that cloud infrastructure IS too. Without that leg, the whole stool falls over. And so if you can guarantee your compute, you can guarantee your storage, but you can’t guarantee your delivery pipe, where do end up? You don’t have a predictable service that a CIO can trust. Whether it’s carriers coming in starting to do N&Is with their private networks to cloud providers or whether it’s enterprises coming directly to Equinix to anchor their regional WAN hubs and using that to connect directly to the clouds.
Lane Patterson 14:24
The first iteration we’ve seen is our program called Direct Connect with AWOS, and that has taken off like crazy. And it involves customers who are co-located at Equinix being able to get to Amazon over a dark fiber cross connect. It’s very simple. If you’re in the same room, why go out over the internet all the way through God-knows-how-many hops and different providers that you have no 800 number to call and so on, and then get back down to your cloud provider who’s on the other side of the room, just go straight. And so for obvious reasons that program has been really successful, but it’s still this kind of nailed up rigid infrastructure of a physical cross connect and physical stuff on each end, and that’s not where services are going. For backbones, people want to build backbones with durable physical infrastructure. But for access to services, that is increasingly virtualized. So people don’t want to consume that necessarily over a dark fiber cross connect, unless you’re talking about gobs and gobs of traffic. So where we see it, that’s a big piece of it is people want virtualized access to those services. But the other piece is people come to a neutral place some they can get access to multiple clouds through the same pipe without going through the internet. That’s the only way to guarantee it.
Jordan Novet 15:50
I think this is going to gain speed. Like the BTIs are going to get some penetration with this model. What I’m trying to figure out also, I write about all these SDN startups, and when I hear Paul Santinelli declaring that SDN is a loser I’m a little confused. Because there are these companies like Niciara, Big Switch they’re getting contracts to get there controllers working on openflow enabled switches. You’re see adoption. Chandra I know you’re familiar with openflow, I’m wondering if this is going to take far longer to be adopted – openflow SDN inside of the data center. I know this is not where you play, but give me a sense, is it loser or are we looking too early?
Lane Patterson 16:39
I know Paul real well by the way, we used to work at Global Crossing back in the day. It may be a loser from a VC angle, from the standpoint of how do you monetize people that want to build controllers and stuff, because it’s pretty open technology, but it’s not a loser from a sense of use cases baked into bigger service concepts. For sure, I think the concept is around and it’s all about the use case.
Jordan Novet 17:01
What do you think?
Chandra Pandey 17:01
As Lane pointed out, if you look into what SDN is bringing, there is a lot of innovation happening in there. The top process is different than what it had been for the last 15, 20 years. So this is not drastically different than how you do things in the box itself. Now you are just extracting it and using the best controller. If each platform had its controller, you could call it like a route processor, switch controller, you could call it whatever you want to call it. And here you are taking it from [inaudible] and putting it in the controller outside, which is much more scalable, you can create a cluster and all that. So the concept is phenomenal, the concept is good. Is it going to be that way at roll out? I’m not sure, I can’t predict the future. But what I see from my point of view, a lot of good things are going to come out of it. And the APIs are going to be much more clear. The devices, the controllers are going to target that interface are going to become a standard. Which is going to enable innovations at very fast clip. And at the end in the industry, it’s like we are blessed with a lot of talented people in our industry, who go and put a lot of effort in and keep on bringing the innovation. But sometimes the innovations kind of just [prottle?] there and you are not able to take advantage of it. And I think this SDN has provided a platform and approach to make those standard interfaces. And then you take advantage of it. Now you can innovate in like minutes, like hour, like days, instead of taking like six months or 12 months for the releases to come out from just like a few networking vendors. So, you kind of open up the platform for the innovations to trickle in. That’s what basically we are looking at. And a lot good of things is going to come out for it. And we have to endorse it and we have to be ready for it. That’s where we want to go.
Jordan Novet 19:11
We’ll see how long it takes. Well guys, thanks so much for taking the time.
Lane Patterson 19:14
Thanks Jordan.
Chandra Pandey 19:15
Yeah.
Jordan Novet 19:15
I appreciate it.

You’re subscribed! If you like, you can update your settings

firstpage of 2

Comments have been disabled for this post