Green Telecom Editor Tony Chan spoke recently to the Joe Weinman, AT&T leader for strategy solutions sales and a well established expert on cloud computing, on what makes the company’s long history, global network and enterprise support capabilities bring to this space.
Green Telecom: So what is AT&T doing the cloud space?
Joe Weinman: The first thing is everyone has a different definition of cloud, and I have this definition that seems to be catching on, and I think it’s helpful. Some people say cloud is IT services delivered across the Internet into a browser. I think that is an overly limited definition of cloud, in fact, it doesn’t represent the use case for most of the enterprises going forward in terms of ways in which they will use the cloud.
So this is what people way, cloud services or IT services accessible over the Internet. There is this thing called private clouds, and that might work with public service provider clouds, everything will ultimately move into the cloud, the economy of scale will drive benefits, etc. I don’t think these are exactly right.
First of all, we think one of the big use cases for enterprise for cloud services is not accessing over the Internet, but accessing through their VPN networks, so their data centre IT resources can burst into the cloud. That is not going to be done over the Internet. So there’s this myth that has been congregated by consumer service providers and our focus, of course, is on the enterprise.
My definition of cloud is: Common, Location-independent, Online, Utility that is available on-Demand – that makes Cloud into a nice acronym.
By common, I mean a shared resource that is statistically multiplexed across multiple customers and applications. Secondly, location independence is key – it shouldn’t matter where the user is, or where the cloud service is. If you use Gmail as an example, you don’t need to know where the data centre is and you should be able to use it anywhere in the world you want to. It’s online means two things – it is accessible over a network and that doesn’t have to mean the Internet or a VPN; and by online it also means it is available, so high availability is key – you need a fall back plan for network access as well as the behind the scene architecture. It’s a utility in terms of the sense that it provides value, but also in terms of usage sensitive pricing based on resources used or other metric of usage. And lastly, it has to be on-demand, so resources should be dynamically allocated.
With this definition, I would claim to you that there are lots of cloud services – electric utilities, rental car chains and hotel chains, because they are ubiquitous, location-independent, I can pay per use per kilowatt, per room day, per car day, or per car mile. Hotels are online because they are accessible by wide area highways and local area hallways. So with that, a lot of people say that there’s a thing called cloud computing and that’s what defines the cloud.
I would say that classic network services have that characteristic, whether it is analogue voice, or VoIP, or data access and transport services – certainly, the typical types of computing infrastructure – computing, storage, data base, etc, coupled with the move into the application stack – horizontal applications like voice, data, video conferencing, and the whole UC type of services. For example, we brought a company called Interwise – about a year and a half ago, and they provided a unified platform over IP. We will sell that software to you, or managed it for you on the premises, but it is also available as a cloud service – you don’t need to know anything, you don’t need to deploy any software, you just go ahead and start using and we will bill you for it.
In addition to that, we have a variety of tools, like managed messaging, hosted contact centres, desktop productivity tools through display over IP or thin client. We have variety of industry standard applications numbering in the dozens that we would provide in a managed, utility, in-the-cloud basis. So we offer managed Exchange, OCS, SharePoint, Oracle Financials, SAP, PeopleSoft, JD Edwards, Siebel, so virtually any enterprise application we will offer in the cloud as a managed app. Lastly, we have some interesting capabilities as the world’s largest network provider, not to mention the world’s largest, not only, telecommunications company, but we’re the world’s largest telecommunications company and/or IT provider. Not many people realize this, but we are larger than Siemens, larger than HP EDS, we are larger than IBM on a revenue basis. If you look at the Fortune 500, the companies that are ahead of us are Walmart, and oil companies and car companies. We are number 10 in the US list, and 29 in the global list.
So we are very much so in cloud services.
We have great global footprint. We are rated at the top in Gartner’s leaders’ quadrant for hosting providers. We are sort of tied for first place – they have mechanisms for ‘ability to execute’ and ‘vision’ – there is no one above us in either dimension. There are a few who are below us in the leaders’ quadrant in either dimension and there are a number that are not in the leaders’ quadrant.
The reason is, we have a decade and a half history starting with managed hosting when the World Wide Web started. Then in early 2000/2001, we started creating utility offers – utility computing and utility storage offers. We combined that with AT&T hosting, with SBC hosting, with Bell South hosting, to create a unified single, integrated organisation which has a foot print of 38 Internet data centres globally.
As you can see, we have four sites in Japan. We just opened our second site in Shanghai last year and we now have a third there. We just opened Bangalore in 2008. Singapore is being expanded – it is a super hub, so we will have additional growth there. The super hub concept has to do with – for those customers who are location agnostic – as opposed to needing to be in a particular country – we may as well aggregate their demand because it just makes sense in terms of the capital expenditure. For all this infrastructure, we have a platform that will provide full, virtual server dynamic bursting in real time. It will allow you to add and remove virtual servers from the cluster. We have grown our utility storage offers. We also have some innovative capabilities on usage-sensitive priced middleware because one of the issues of cloud computing is, just getting the infrastructure on a pay per usage basis doesn’t matter if the software has to be an outright license purchase. That has driven people to open source, which we certainly offer.
I would claim that, with both the industry definition of cloud as well as with my broader definition of cloud, that not only are we in the cloud business, I would claim that we are the number one cloud provider globally for the enterprise. Now, as far as consumer services, there are already vendors there and they are trying to move into the enterprise. But AT&T is what I called an integrated company, and there aren’t many of those. There are companies that are good at networking. There are companies who are good at hosting. We are excellent in networking – we are in the leaders’ quadrant for US networking, for pan-European networking, for Asia Pacific networking and for global networking, we are basically in first place, or tied for first place in each of those. In addition to that, we have this leadership position in hosting. When you combined those, who else is there? There are some companies that have a hosting presence, but they don’t run networks.
What is this important?
I believe that being an integrated provider is absolutely key for reasons that have to do with cost and performance, especially if we look at the current, emerging generation of Web applications. For example, if you considered video over the Web, latency is absolutely key. Ajax application – same thing – you’ve got 150 to 185 millisecond maximum time budget, round trip, including server time, to be able to deliver that. So you can’t afford extra hops to go offnet to a third party provider and extra hops back. So we have created this incredible infrastructure that ties together the world’s largest global network – we carry more IP traffic than anybody – 16.5 Petabytes per day, and growing, 88 subsea cable systems, 880,000 route miles of fibre – an as you know every route mile can be multiple fibre pairs, and anyone of those fibres can carry up to 80 service channels, which use to be 2.5Gbps, then they were 10Gbps and now they are 40Gbps – we have been buying the majority of the 40Gbps (OC-768) line cards manufactured in the world – so we know we are ahead of anyone in terms of those deployments. Not to mention we have a very aggressive 3G roll out plan. We have already deployed in hundreds of markets. We have almost 100,000 Wi-Fi hotspots globally.
We combine cloud computing, whether it is infrastructure, platform, or applications services that we have on the footprint, and couple it with the network, coupled with enterprise sales capability, the ability to operate at scale, high availability, and the enterprise SLAs.
Without being overly arrogant, the consumer-grade services have a different approach to service roll out and availability – everything is in beta, possibility still now even after five years. Beta use to mean something – this is a step on the road to testing before we provide a quality product that we will stand behind. Beta now is just a disclaimer – if it doesn’t work then too bad, you are on your own. And you know as well as I do the recent outages in a storage services that lasted for days, and a mail service that went down. Our customers don’t live with that, they don’t accept that. In the global enterprises that we serve, it’s three 9s, four 9s, or five 9s availability.
And any outages, on the odd chance that they happened, are dealt with on an advanced notification and escalation process, a multiple hundred million dollar global network operations centre, multiple customer support centres and teams. Selling to the enterprise, and more importantly, delivering the quality services to the enterprise for mission-critical services for large global enterprises is a very different value proposition then offering something for free that is kind of on a best effort basis.
So I would claim that we are not just in the cloud field, but we are – what the lawyers would let me say – a leading provider.
How is AT&T’s cloud offering different from others?
We will do things like sell capacity, again they can be dynamically allocated in a variety of different formats – virtual servers, physical servers, and we can do virtual servers essentially instantaneously. In addition to that, we have what we called managed hosted applications that are available on a pay per use basis. There, the cloud concept is a little bit different in a way – it is not the same type of infrastructure bursting that you’d expect because you are dealing with a lot of large numbers around statistical aggregation of users. We can allocate dynamically, in fact, we run hosted SAP on our Synaptic Hosting platform, so what that means is we can scale out the SAP instance by leveraging the inherent, dynamic capabilities of our underlying Synaptic Hosting platform.
That is interesting because it has almost got two dimensions of flexibility – there’s the pay per use around the users, and there’s the flexibility of scaling based on use. So there’s a variety of scenario for that – both for business as usual variation but also for business continuity and disaster recovery, where a lot of customers may still want to have SAP running in their own data centre, but are saying ‘why do we want to run a separate data centre with an idle instance, why not leverage AT&T, who is a preferred partner of SAP, together with the Synaptic Hosting platform, so that in the event of an outage, we can fail over to an AT&T facility.
A lot of this stuff is applicable to variable and unpredictable demand. A really good example that we did was providing support for the US Olympic Committee. Think about that as a cloud spiking problem. You only have two hours of demand – its two weeks, but during the opening ceremony, you went from nothing to all of the sudden, people went crazy, and there was unprecedented demand. We were able to provision their site for them in our Synaptic Hosting facilities within a very short amount of time and then burst to meet their unpredictable demand, not just for hosting services, but also content delivery because they were doing things like streaming video. We do the Grammy awards, American Idol, where we broke all of our text messaging – which is also a cloud service – records actually for this most recent season of American Idol. There was something like 77 million messages sent because people texted each other and they vote.
There’s two ways you can take what I’m saying – you can say ‘alright, this guy has a good story on Synaptic Hosting, every once in a while he starts to talk about text messaging, maybe he is too dumb, he doesn’t quite get that, that is different than what I have in mind when I think about cloud services.’
But if you think about what’s really going on, you’ve got this convergence of different communications, social networking, Web-oriented, mobile Web, coming together with mobile access networks, coming together with Wi-Fi as an alternative distribution path into mobile devices, or for that matter, using Femto cell technology, Wi-Fi networks become important, coupled with wired line, coupled with remote VPN – you can’t really break out any longer, ‘oh, this is a Web application, you are talking about a voice application’, it’s all coming together. You’ve got click to chat, you’ve got VoIP handling communications to people who are both online using a softphone on their PC or an integrated UC portal.
So, we believe that having all the capabilities that we have through our AT&T Labs heritage, coupled with the network and hosting, just really brings together this nice package. You add in digital media solutions, add in content delivery and content management platforms, and it says that we are in the best position, both to provide point solutions but I think it gets more interesting when you think about integrated solution that start to say – you don’t care what device you are using, you don’t care what network you are coming in on, but what you do care about is user experience, the richness of that experience, the latency, the availability and the ability to do multi-mode – you want immersive video – great, you want Webcam video-great, you want mobile video-great, you want voice, you want VoIP, you want data, you want whiteboarding, you want to accessing it all over a Web platform, leveraging mobile device capabilities like accelerometers, touch interfaces that are not yet built into the browser paradigm. We can weave all that together.
How important is the network in the cloud computing service model?
If I only sold coffee cups, I would say that coffee cups were the most important model. If I was a department store that sold lots of things, coffee cups, nails, and cards and everything else, then I would say, ‘why are coffee cups the only way to go?’ So in terms of our breath of services, we not only offer just compute services – we do all that – but we do much more than that. So why should we limit things and say ‘well, we only sell coffee cups.’
Secondly, where some of these companies that started out with only one data centre, they are playing catch up, and they are now trying to expand into other parts of the world. We are already there. You can count the countries – we are in a dozen and a half countries.
So what we have is large data centres in all of these areas and now we can do one of two models. One is if you only want to be in a single data centre as a customer, we can do that for you. And there may be solid reasons for that based on your current size, based on compliance requirements. If you want to be in multiple locations, we have a variety of patented technology that leverages location based sensing together with optimum routing to the nearest node. So we will do global load balancing across, say, web servers where ever you tell us to deploy them. We can deploy them, scale them dynamically. We have something called adaptive any stream technology that is being rolled out first half of this year. What that does is provide optimised routing to the nearest service node based on location identification.
So again, because we run the network and we know where you are coming in in the network, therefore, we can optimally route you, which reduces cost, reduces latency, which in turn, enhances user experience. So between, flexibility through Synaptic Hosting, optimised routing through IDNS, or intelligent DNS services, together with http direct and global load balancing, we can optimised the whole experience across this disperse architecture, which we believe, represents the ultimate in a compute architecture.
Tightly coupled network and compute fabric, geographically dispersed, serving a global user based, which can be employees, customers, customers’ business partners, with optimised routing to a service node, as well as failover through MPLS fast reroute on our global meshed network, which is lowest latency, highly reliable – again, 88 subsea cable systems gives you a very robust mesh.
And we not only do optimised routing in that mesh, for example, we are deploying cable systems that avoid the Taiwan off shore earthquake epicentre. In addition to that, we have enough cable systems going in and alternate routes – North Pacific, South Pacific, we just announced the Europe India Gateway, so we’ve got alternate paths. So no matter where in the world you are and no matter where you want to get to, we will provide you with multiple alternate paths that are also high capacity.
We have also just completed a major, global router upgrade, so we are deploying the state-of-the-art Cisco CRS-1s, which are native 40Gbps routers with a high availability, stateful failover routing architecture and a 96Tbps backpane. We’ve put dozens in, both in the US and globally.