NGN Notes

I went to NGN2002 expecting the worst – yet another event better named No Good News than Next-Gen Networks – but after a full week of sessions that illustrated the continued vitality of technological development in the face of diminishing markets, I’d say there’s hope. The best minds in the business are hanging in there and figuring out how to fix this mess, survive this cataclysm, and continue to engage in the evolution of the telecom network.

In recent weeks I’d been feeling extremely negative on telecom, seeing no signs of a recovery, at least not till 2005, which is far enough in the future to render futile any attempted divination of the when and how of spending resurgence.

The wisest of the people we talk with have stopped asking when the recovery will occur and instead ask what are the catalysts for recovery – guideposts to watch for in the coming years that signal carrier spending is about to tick upwards, with a focus on next-gen gear. Regulatory shifts, elimination of carrier debt loads, flushing out of equipment inventories, and reaching a critical mass of broadband homes in the U.S. are the most popular catalysts named; but it’s worth arguing that carrier decisions on network architectures – MPLS core or more ATM? RPR, Sonet or native Ethernet metros? 802.11 or 3G? – represent an even more fundamental catalyst for next-gen networking recovery.

NGN, it turned out, was a place to help determine just how emerging technologies are maturing and what carriers are saying about them.

Encouragingly, plenty of carriers showed up. I saw AT&T Corp. (NYSE: T) badges, Bell Canada (NYSE/Toronto: BCE), a few European PTTs, BellSouth Corp. (NYSE: BLS), and a number of CLEC representatives (with travel budgets!). They asked lots of questions during the sessions and talked animatedly in the hallways and receptions. None of them said, “Hey, I’m upping my spending by a billion next year, this MPLS stuff is ready for prime time!” But the level of engagement on topics such as wireless LAN, IP QOS, Internet security, VPNs, and all things Ethernet was intense.

I missed most of the keynotes, but when I poked my head in, there wasn’t much being said that hasn’t been repeated all year (and inviting a security guru from the White House was a little too much evidence of our national transformation into a police state, so I skipped that as well). The opening keynote, Public Networking, The Worst Year in History, from the conference co-chairmen – David Passmore, Research Director of the Burton Group, and Dr. John M. McQuillan, President of McQuillan Ventures – was probably not the best way to keep attendees listening for five days, but amid the mephitic malaise they did note some important trends:

My colleague Rick Thompson was there with me; what follows are our notes on the various sessions, as well as what we heard in the hallways and “Birds of a Feather” sessions after hours.

Here’s a hyperlinked summary:

— Scott Clavenna is Director of Research at Light Reading and President of PointEast Research LLC; Rick Thompson is Principal at PointEast Research.

Want to know more? The big cheeses of the optical networking industry will be discussing Next-Generation Networking at Lightspeed Europe. Check it out at Lightspeed Europe 02.

Metro

In my session on Metro RPR vs. Next-Gen Sonet vs. Ethernet, I asked for a show of hands on which of the three solutions would prevail in five-years time. Of a crowd of 150, native Ethernet and next-gen Sonet got about 60 votes each, RPR… three (3). (For more on these technologies, check out Next-Gen Sonet , Metro Ethernet and Resilient Packet Ring Technology).

Lots of talk in the hallways about GFP (generic framing procedure). This is to be watched, as it represents the next step in the evolution of Sonet. It elegantly maps a broad range of protocols into virtually concatenated Sonet/SDH channels, without protocol translations, making possible what many ILECs and PTTs are asking for today: a unified metro edge that supports TDM, Ethernet, and storage over a common transmission platform.

Long-Haul Optical

This was set up to struggle. They put this session last in the conference, during what would typically be lunch – as if these guys didn’t have it rough enough already. Presenters from Nortel Networks Corp. (NYSE/Toronto: NT), PhotonEx Corp., Ciena Corp. (Nasdaq: CIEN), and Mintera Corp. were more alike than dissimilar on this occasion: They need spending to return, plain and simple. The debates over whether to deploy 40G vs. more 10G are somewhat irrelevant until carriers start expanding their cores.

When asked what was the most important component innovation they looked forward to, almost all mentioned active dispersion compensation, particularly electronic solutions, since costs are expected to be much lower than optical solutions. Others talked about ROADMs (remote optical add/drop multiplexers), and tunables (lasers, receivers); but the focus remained on expanding backbones with the lowest possible capex.

Optical Subsystems and Components

Plenty of talk about subsystems, from intelligent amplifiers, to tunable dispersion compensation solutions, to monitoring and tunable transponders. The interesting comment here came from Charles Corbalis, president of Calient Networks Inc. He described the history of switched services, from T1s to T3s, to OCxs, and proposed that the stage was set for the proliferation of wavelength services.

Why? Because these switched TDM services arose once carriers first proved them in the trunks of the network, then built private networks for large enterprise customers, then finally rolled them out as a common service. Over the past few years, DWDM has been deployed as a transport solution in the core; more recently, large enterprises have built private DWDM backbones; so logic would dictate that switched wave services are on their way.

Could be, but my only concern would be that these switched TDM services leveraged developments in silicon and digital networking, whereas DWDM services remain in the analog domain, making them more difficult to scale, monitor, and provision.

It’s clear that innovation and funding has followed the tide in from “core” to “edge” to “access.” Now that many of the access players are well underway, from a development and funding perspective anyway (next-gen Digital Loop Carrier, Ethernet over Copper, Ethernet over Fiber, Optical Access, Wireless Access, etc.), it seems as though the innovation tide is starting to move back out, slightly, toward a new definition of edge networking.

In the first go-round, we saw many developments at both the service layer and transport layer.

Service layer innovation was largely characterized by scaleable edge routers and IP service switches integrating multiple data service functions into complex, software-intensive, ASIC-intensive platforms. One or two of the players will find success, while the rest will struggle and ultimately shut down.

Transport layer innovation was largely characterized by God boxes trying to over-integrate data and TDM functionality into the same platform. Again, a few made it, but ended up becoming next-gen Sonet boxes. The remainder struggle to reposition, as the market remains cold and the technology becomes stale.

It’s likely that access technologies will continue to be purpose-built for specific applications, services, types of customers, geographies, etc. It’s also likely that carrier uncertainty will remain, relative to defining and delivering quality IP services and effectively managing data and TDM integration. This will force the new edge of the network to be manageably flexible in nature, supporting multiple access technologies and adapting to fluid service demands. Probably sounds a bit idealistic, but it seems as if steps in that direction are looking fairly realistic:

What’s driving the innovation of third-generation edge networking?

In closing on this topic, a quick reference to AT&T’s keynote presentation, The Network Evolution: Journey to a Multi-Service Edge, Self-Operating Network, delivered by Dr. Hossein Eslambolchi, CTO: The short story here echoed many of the points above relative to the flexibility at the edge still being a requirement and a priority for AT&T going forward. It was a bit idealistic in its view of self-operating and self-provisioning for any service. Perhaps somewhere between the hoped-for inherently intelligent network and today’s inflexible infrastructure, a pragmatic mix of easily taught or programmable platforms will arise.

There were several “Birds of a Feather” sessions held on Wednesday evening. These amounted to small, interactive tutorials on specific technologies. Some interesting ideas came from the grid computing session (see Watch for the Grid). It started with a basic discussion of grid computing, which in reality has many definitions, but basically deals with interconnecting multiple, distributed servers to create a collaborative, high-performance computing resource. Not being a supercomputer expert by any means, I tend to think of the networking infrastructure aspects and opportunities. Below are some notes on the basic layers of grid computing/networking, and some thoughts on integration.

Basic functional layers as defined in the session:

Today, different companies deliver solutions for each specific layer. The application layer will probably always be that way, but it seems there may be an opportunity to integrate the bottom three layers (serviceware, middleware, and fabric) into a specific grid switch type of solution. The infrastructure solution could be a product for both grid application service providers and grid resource providers, the primary difference being: The former manages everything, including the application; while the latter simply provides the infrastructure from which multiple applications may be delivered.

Today’s layer-specific solutions (e.g., Avaki Corp. and Entropia Inc. – middleware layer; Force10 Networks Inc. – fabric layer) are more than adequate, as the market is in very early stages, from both the research angle and the Web services angle. As the market develops, the integrated solution can cut costs and better enable the proliferation of the types of providers mentioned above. Sun Microsystems Inc. (Nasdaq: SUNW) and IBM Corp. (NYSE: IBM) are aggressively pursuing grid computing, and Cisco Systems Inc. (Nasdaq: CSCO) is involved in ongoing investigation as well.

A few other tidbits on this topic include some of the following points:

It seemed as if nearly half of the sessions involved WiFi, 3G, wireless security, and so on. The recent lack of interest in traditional wireline telecom networking served to put wireless in the show spotlight – not a great surprise, I guess. There was, though, a significant amount of chatter around the fact that many of the wireless technologies are also over-invested (see Funding: Startup Roundup). With confusion building around 3G and 802.11x – as to whether the technologies are competitive, complementary, or both – there certainly seems to be some opportunity for future positioning.

It’s clear that the rapid deployment of 802.11 technologies has created opportunities for startups, challenges for users (and startups), and questions or concerns for many of the cellular players (vendors and service providers). Aside from the various tutorial sessions on cellular (2G, 2.5G, 3G technology choices, architectures) and WiFi (802.11b, 802.11a, 802.11g), there were a few good sessions on coexistence and/or competition of WiFi and cellular (some notes below). Also, there is no dispute that security is a primary area of concern for WiFi going forward, and there were a number of sessions that focused on that.

* * * * *

One of the sessions included panelists from Airvana Inc., a startup working on next-gen 3G infrastructure equipment based on 1xEV-DO (optimized for packet data, in contrast to 1xEV-DV optimized for data and voice), and YAS Broadband Ventures, a VC presenting from more of a WiFi perspective. In addition to the standard security concerns, some of the biggest service provider concerns about WiFi, from the cellular view, include high backhaul costs due to the sheer number of sites, site rental and maintenance, and long-term congestion control. Two other interesting points were:

The view from YAS Broadband is that simple applications like email and voice will be adequately handled via current next-generation cellular technologies, but that video will not. Security can and will be managed via many of the same mechanisms that exist in cellular networks; although much of the same security can be applied to 802.11 microcells, the challenge will be in maintaining security and connection when roaming between cellular and WiFi networks.

This seems to be the one common theme from both camps. The fact is that both technologies will exist, but integration may hold the most promise for developing new solutions. Integration could mean integrating WiFi and 3G into the same device, which is possible, but has tradeoffs such as cost, size, and power consumption (classic integration issues). The more realistic type of integration is probably managing the roaming between the two different networks from various perspectives, including authentication, billing, and connection maintenance.

At the end of the day, I would say, the most agreed upon conclusion was that 802.11 and 2.5/3G would co-exist, with slight momentum behind the thought that if one were to overtake the other it would be 802.11. A final thought is: Even if the technology permits that to happen, it may be political forces that prevent it.

* * * * *

The other interesting wireless session included panelists from Aruba Networks Inc., a startup working on a next-gen wireless LAN switch; and MeshNetworks Inc., a startup working on ad hoc networking.

Aruba presented today’s enterprise network in three layers:

Aruba seems to be of the opinion that access points will continue to be driven down in price, but WLAN functionality requirements will continue to increase – in managing 802.11b/a/g scaleability, RF visibility, secure enterprise-wide roaming, and upgradeable security (changing encryption schemes). The model above seems to suggest that the core remains wired with high-performance enterprise gear, but that the distribution grows with highly functional WLAN switches, fed by cheaper and cheaper access points.

Mesh Networks presented from more of a metro/wide-area perspective, adding to the debate between 802.11 and cellular. They’re using ad hoc networking technologies to solve coverage problems that are more prevalent for emerging 802.11 than for today’s cellular. To solve the issue of coverage they use multi-hop mesh networking to extend the coverage of 802.11. The ad hoc approach inherently suggests that the network gets stronger with the addition of each node (essentially, each node provides bandwidth for the overall network). So not only does it potentially extend range, but it also increases network capacity and scaleability as it reduces interference.

The multi-hop process can also potentially counteract increased access point cost (as security is added). Mesh expects a negligible 1ms delay per hop, with support for 30 to 40 hops (with 802.11b, 802.11a unknown). A lot of these metrics are dictated by the routing algorithms. In many ways, this is similar to applying intelligent routing to circuit-switched networks (as with GMPLS), in that it is constraint-based. In optical networks, the constraints could include trunk size or switch granularity. In wireless ad hoc networks the constraints could be, say, available capacity or power characteristics of intermediate nodes.

One other issue, is that because all nodes may be used as intermediate access points, security could become more complex than simply taking care of primary access points. This might further complicate the routing algorithms required.

* * * * *

Other random wireless musings:

On Wednesday, October 16, there was a venture capitalist panel discussion titled After the Party’s Over, Investing in Networking Today. The panelists included representatives from ComVentures, Integral Capital Partners, and Morgenthaler. Below are some notes from this session, which centered around three primary topics.