AT&T’s New Gods

Hossein Eslambolchi, AT&T Corp.’s (NYSE: T) CTO, is getting giddy about MPLS. And AT&T is about to make one of the largest commitments to the technology.

At the Network Outlook conference here in San Francisco this week, Eslambolchi elaborated on the “Concept of One,” and then on Thursday on a conference call, he and Mike Jenner, VP of AT&T Global IP Network Services, spent over an hour talking Multiprotocol Label Switching (MPLS) and how it would unify the entire AT&T network.

In a span of three short years, AT&T believes it will go from today’s network of many to tomorrow’s network of one, using MPLS as the glue to unify all services, all customer applications, and all customer interaction with the network within a single logical framework – all by 2005. Is that possible? Not in three years. Is it necessary? Let’s see.

Eslambolchi and Jenner claim their enterprise customers are clamoring for a more unified network services offering, as they transform their own networks by unifying around Internet Protocol (IP). Under AT&T’s Concept of One, legacy networks such as Asynchronous Transfer Mode (ATM) and frame relay will be consolidated logically into a single network infrastructure through the implementation of MPLS and interworking standards. Already, customers of AT&T’s frame relay service can transition to “IP-enabled Frame” without any new gear at the customer premises. Their service is transitioned within the POP from the core ATM network to a core MPLS network, compliments of an MPLS blade on Cisco’s ATM switches. Any-to-any connectivity is supported, and the process for scaling any frame relay service is much simplified.

Concept of One is extended now to encompass all access services, which will someday be aggregated onto a new edge device called the MSP (multiservice access platform), a sort of “mini-God box” that accepts all sorts of services, packetizes them, and then aggregates them onto a Gigabit Ethernet link for transport to a larger office where a box called the MSE (Multiservice Edge) will aggregate those further and hand them off to the switched MPLS core. Both of these “mini-God” and “Big-God” boxes are interesting to behold. Think of the mini-God as a low-cost pizza box that can sit at the customer premises or end office, and the other, larger God as a very scalable edge platform that can handle up to 400 Gbit/s and provide all the necessary interworking between the legacy AT&T networks and the new MPLS-enabled core.

This undertaking, dubbed “Project Pluto” over the past few years, is coming to a head. AT&T claims they will make a vendor selection on the MSE by the end of the month, but every vendor I’ve spoken with says the requirements are impossible to fulfill, so the selection will pretty much involve the definition of what AT&T wants followed by a forced partnership between a startup and an OEM to accomplish the task. Who supplies the pizza box (and can make a profit on it) remains an open question.

Isn’t it funny how everyone hacks on startups for trying to build God boxes (only after they have been spurred to do so by an RFP) that, of course, no incumbent vendors would ever use? They ask you to build a God box and when you do you are ridiculed for your hubris.

But, before looking forward, let’s first look back. The first step in the Concept of One is to retire 130 legacy systems, though AT&T has not specified which ones. I would guess that means Cisco Systems Inc. (Nasdaq: CSCO) (Stratacom) ATM switches, but it’s impossible to say until that happens. These boxes will be replaced with whatever is selected for the MSE, a system with plenty of legacy interfaces facing out and IP/MPLS facing into the network core. Which are the leading contenders? I’m guessing they’re most likely Cisco, Alcatel SA (NYSE: ALA; Paris: CGEP:PA), and startups like Équipe Communications Corp. and Laurel Networks Inc. Cisco must be considered the leader, because the AT&T data network is primarily Cisco gear.

With the MSE boxes in place, AT&T can MPLS-enable all of their network services, from frame relay to ATM to IP. VPNs will now be MPLS VPNs, and all the legacy services will be “IP-enabled.” The beauty part, AT&T claims, is the new level of consistency at the edge of the network. Performance measurement, management, and SLA verification will all be standardized around MPLS – though the AT&T execs we spoke to were clear in saying this is AT&T’s MPLS, so this functionality applies to those services beginning and ending on AT&T’s network.

Beyond the Concept of One, which intends to unify and integrate network services around an infrastructure of Ethernet transport and MPLS-based service creation, there is the “Concept of Zero,” in which service delivery is both highly reliable and automated. Within the Concept of Zero, all human-to-human and human-to-machine interaction eventually becomes automated, putting application and service control in the hands of the customer rather than the network operator.

Eslambolchi claims this is the first time in AT&T’s history that it is actively opening its network to the customer, enabling new levels of customer network management, service creation, and ordering. This is largely accomplished through new OSS (operations support systems) and BSS (business support systems) developments and is aimed at empowering the enterprise customers with the tools to basically create their own network services as they transform their own internal networks.

In this light, Eslambolchi sees AT&T’s future network as one based on two functional layers: one of connectivity, another of mediation. Mediation, he claims, is the process of bringing applications onto the unified network infrastructure and includes the building blocks of routing, security, directories, and connection control. A service like voice is no longer conceptualized as a service, per se, but an application that rides on a distributed IP network. Same goes for multicast video, corporate extranets, e-learning, and garden variety enterprise apps like CRM (customer relationship management). Application servers are accessed across the network by any variety of devices imaginable, from desktop PCs to supercomputers, mainframes, PDAs, cell phones, and televisions.

This is the heart of the matter, and reinforces what another large carrier recently told me. The edge of the network must become adaptive and programmable for application support. If AT&T’s Concept of One pulls that off, Eslambolchi deserves sainthood. If not, well… They have made some nice looking PowerPoint.

— Scott Clavenna, Director of Research, Light Reading

<<   <   Page 2 / 3   >   >>
gea 12/5/2012 | 12:25:30 AM
re: AT&T’s New Gods Damn this is the kind of thing that makes the geek in me get all hot and bothered. Don't tell me it won't work or that it'll cost far more than it will ever make, 'cause I'm not trying to hear that, see?

If it worked (and I've got my fingers in my ears so yes it will yes it will yes it will), one could imagine the existence of intermediate companies that create specialized services on top of the AT&T MPLS infrastructure and then sell them to end users (whereas AT&T's own apps might be more wholesale/Large Market-oriented).
road__runner 12/5/2012 | 12:25:29 AM
re: AT&T’s New Gods Can't that be done today ? It seems like a customer could buy an ATM PVC/SVC from AT&T today and sell their own services on top of it. Or are there contractual clauses that prevent customers from doing that ?
eschrob 12/5/2012 | 12:25:28 AM
re: AT&T’s New Gods Scott- I've seen so many TLAs that I numb. Can you provide a writeup or link that better explains these?
MSP- MultiService Access Platform
MSSP- MultiService Switching Platform
MSPP- MultiService Provisioning Platform
MSE- MultiService Edge ___?

How do these differ from Routers?
jim_smith 12/5/2012 | 12:25:27 AM
re: AT&T’s New Gods "... one could imagine the existence of intermediate companies that create specialized services on top of the AT&T MPLS infrastructure and then sell them to end users ..."


Creating specialized services over leased bandwidth is not a new business model. ISPs and wireless service providers do that today.

The only new thing I see with an (G)MPLS infrastructure is the "bandwidth on demand" capability. It will be interesting to see whether the "bandwidth on demand" capability will really create new services, or it will die just like the other penniless cool ideas that have seen in the recent past.
Scott Clavenna 12/5/2012 | 12:25:24 AM
re: AT&T’s New Gods Here's a shot, though some of these are AT&T specific, as noted.

MSP- MultiService Access Platform
This is what AT&T, or at least the CTO, has conceptualized as the universal access platform, meaning it is small (1 rack unit), supports both circuit and packet client side interfaces, then packetizes those and aggregates them into a single Gigabit Ethernet pipe for carriage upstream. This is a fiber fed device, so not exactly universal, but would be located at the business customer prem, and maybe the home someday once it achieves commodity level pricing. Today, these are designed to cost around $3,000 to $5,000 each.

MSSP- MultiService Switching Platform
This, I would guess though I haven't seen this one very often, refers to the ATM-brand of multiservice switch, meaning a cell-based switching system that adapts all traffic to ATM cells and transports them across the WAN. This could one day be replaced by an MPLS-based switching system, AT&T's MSE, which would append an MPLS label to all packetized traffic flows and transport them across an MPLS-enabled WAN.

MSPP- MultiService Provisioning Platform
This has commonly been used to refer to those systems that take multiple services in (TDM, ATM, Ethernet) and map them into SONET using standardized (GFP, LAPS) or proprietary mapping/framing mechanisms. These systems tend to have both packet switch fabrics and TDM cross connects, so perform some level of packet aggregation and switching plus bandiwdth management.

Since this is SONET oriented, it typically resides within the transport network, rather than within the data network and is more geared around metro services rather than WAN services. AT&T today uses the Cisco 15454 as their MSPP but has claimed they want to get greater efficiencies out of this platform by migrating to a pure-packet solution based on MPLS and Ethernet, the MSP.

MSE- MultiService Edge ___?
This is a term used by AT&T to describe the evolution of their ATM switching infrastructure, today provided primarily by Cisco, through their acquisition of Stratacom years ago. This new MSE box would take in ATM, frame relay, Ethernet and IP but would no longer be based on a cell-switch fabric. Instead, it would be a hybrid of sorts, turning every service into MPLS and providing the necessary interworking between these layer 2 services and the MPLS-based core IP network already deployed in AT&T. This interworking is key at the control plane level, so end devices can use their existing operations methods but at the same time take advantage of the scalability and ultimate ubiquity of MPLS in the core. This is essential once you envision the millions of end points on a large network like this. ATM struggles to scale cost-effectively and operationally to those levels.

How do these differ from Routers?

The boundaries are blurring, but basically these devices have much richer layer 2 capabilities and a much wider range of client side interfaces. But again, the boundaries are blurring, and many router vendors believe they can be designed as MSEs or MSPs through some control plane tweaking and new line cards, so you'll see lots of ongoing debate and confusion here, to be sure.

gea 12/5/2012 | 12:25:09 AM
re: AT&T’s New Gods "Creating specialized services over leased bandwidth is not a new business model. ISPs and wireless service providers do that today."

Agreed, but it's different from what could be possible with a full-fledged MPLS infrastructure. Current specialized services are still circuit-based, even where ATM exists. (ATM does not support best effort.)

And if this infrastrcture were end-to-end, then one day we might actually see something like "Video Conferencing on demand" (where a temporary, high-bandwidth real time flow is created for an hour or two), or currently unimaginable services.

Needless to say, this has been a sort of holy grail in both ATM as well as MPLS circles, but it's cool to see some giant carrier with tons of $$$ actually try it (of course, I'm not an AT&T investor....)

MPLS guy 12/5/2012 | 12:24:55 AM
re: AT&T’s New Gods Laurie,

Nice tip.

With Masergy running MPLS, over Native IP, the future is already here.



rush21 12/5/2012 | 12:24:47 AM
re: AT&T’s New Gods
Despite MPLS guy's unabashed "homerism" and love for all things Masergy, he is right about one thing. Masergy is already doing exactly what AT&T is hoping to do in three years. MPLS core, MPLS Layer 2 and 3 VPNs, Ethernet transport, Frame-relay replacement, automated provisioning, end-to-end QoS (based on TOS and not CIR), and customer control/automated service creation. Being able to start from scratch has its advantages!
fiber_r_us 12/5/2012 | 12:24:45 AM
re: AT&T’s New Gods Level 3 has been doing this for over two years now...
fundamental_guy 12/5/2012 | 12:24:27 AM
re: AT&T’s New Gods That explains their problems.
<<   <   Page 2 / 3   >   >>
Sign In