AlcaLu Joins the War for the Optical Core

Alcatel-Lucent (NYSE: ALU) unveiled its next-generation optical crossconnect today, fueling what's looking to be a battle for future core-network designs that combine packet and optical functionality.

The optical core hasn't seen much dramatic change for nearly a decade, but the convergence of packet and optical networks -- along with the expected flood of Internet video traffic -- is giving equipment vendors the motivation to revamp the core.

AlcaLu presaged the 1870 Transport Tera Switch, as it's called, in September, when it announced its plan for "converged backbone transformation" and what it calls "high-leverage networks" -- ideas that hinge on mixing packet and optical capabilities into one box. (See AlcaLu Preps Grand Convergence Plan and AlcaLu Makes Its Packet-Optical Move .) (And here we thought they were just making up new terms for fun.)

Alcatel-Lucent particularly likes this plan because, unlike many of its competitors, it's got deep expertise on both the optical side and the packet side. "Given the level of integration and the complexity of the systems, this is a place where we have a competitive advantage," says Alberto Valsecchi, vice president of marketing.

The core network has already seen some next-generation boxes, including: The Ciena Corp. (NYSE: CIEN) 5400 family; the Z77 from startup Cyan Optics Inc. ; the Fujitsu Network Communications Inc. Flashwave 9500; and the Huawei Technologies Co. Ltd. OSN 8800 (which so far lacks packet capabilities but has a next-generation kind of capacity). (See Ciena Catches Packet/Optical Convergence Bug, Cyan Plays God With Optical, and Huawei Intros Big Crossconnect .)

But the 1870 is arguably the closest box yet to a truly next-generation packet/optical core switch. "They're at the front of the whole, for lack of a better term, God Box that combines the Layer 2 and some of the Layer 0 functionality," says analyst Andrew Schmitt of Infonetics Research Inc.

The 1870 certainly has the feel of a God box. It handles Ethernet, Sonet/SDH, and the Optical Transport Network (OTN) all within one fabric, a chip Alcatel-Lucent has worked on for years. It's also got grooming capabilities down to the ODU0 level -- OTN's equivalent of a Gigabit Ethernet channel. And it can be outfitted with reconfigurable optical add/drop multiplexer (ROADM) cards.

It also happens to be just plain big, with capacity of 4 Tbit/s going into card slots that can handle 120 Gbit/s each. (An upgrade to the box will double both those numbers later, AlcaLu says.)

If only it had DWDM capabilities, the 1870 would truly match the Heavy Reading definition of a core packet-optical transport system (P-OTS). (See Redefining P-OTS.)

"In the camp of packet/optical, AlcaLu's got really good solutions," Schmitt says. "They were working on that since I was at Vitesse Semiconductor Corp. (Nasdaq: VTSS) back in 2003. They've got a huge investment in that platform that no one can just show up and replicate."

What makes this kind of box important is that it fits some carriers' visions for the future network. Most big carriers want to see the network's packet and optical sides merge, so that they're controlled by the same boxes and the same management systems.

This would simplify the network. In some scenarios, it would also save money by replacing expensive core-router ports with optical transport. That's the plan Verizon Communications Inc. (NYSE: VZ) is chasing, having issued a request for proposals around a converged P-OTS. (See Verizon Rethinks Long Haul.)

The 1870 seems to fit the requirements, and AlcaLu officials describe the system in Verizon-friendly terms. Specifically, the 1870 is able to switch traffic at the lowest possible layer of the network. That way, traffic that's just passing through a node can be switched optically, without having to be terminated -- avoiding the need for those pesky core-router ports.

What's interesting is that AlcaLu isn't putting all its eggs in a Verizon-like basket. The company has also worked on IP-over-DWDM, a technology that puts optical capabilities onto the router (as opposed to adding packet smarts to optical switches). Not surprisingly, that approach is being pushed hardest by the router vendors -- Juniper Networks Inc. (NYSE: JNPR), in partnership with Nokia Siemens, and, of course, Cisco Systems Inc. (Nasdaq: CSCO).

"The bigger vendors are hedging. You see the same thing with Nokia Siemens," says Heavy Reading analyst Sterling Perrin.

That's wise, because each camp is likely to draw support from some major carriers. The eventual outcome will likely be a spectrum of deployments that include shades of gray between the two approaches, Perrin says.

That's why these systems are all likely to end up looking like God boxes. But God boxes were built on the assumption that carriers would use all of those features -- specifically, that they'd be switching every type of traffic through these boxes -- and, in that regard, God is dead.

"I don't think anybody wants to use everything these boxes do. The modular approach is going to be a big deal," Perrin says.

For example, while all the new core boxes will support OTN, many networks aren't likely to switch to OTN for years. That's why the 1870 and its ilk will continue to support carrier Ethernet and good old Sonet/SDH.

— Craig Matsumoto, West Coast Editor, Light Reading

mvissers 12/5/2012 | 4:44:19 PM
re: AlcaLu Joins the War for the Optical Core

ODUflex is being added to the next version of OTN line port devices. Once those new line ports are added to the OTN systems the network is capable to support switched ODUflex connections.

Note that the new OTU4 and OTU3e2 line ports treat ODUk (k=0,1,2,2e,3) signals as specific bit rate versions of the generic ODUflex signals; i.e. all those ODUk/ODUflex signals are mapped into the HO ODU4 and HO ODU3e2 by means of the new Generic Mapping Procedure (GMP). GMP is a generalized asynchronous mapping technology that is agnostic of the bit rates of the client and server signals. You only need to configure the bit rate and tolerance of the LO ODU signal and the tributary slots in which it is transported in the line port to have the GMP process perform its task to map/demap this LO ODU signal into/from the HO ODU tributary slots. This programmability also enables the hitless resizing of the ODUflex connections when such ODUflex connection carries GFP-F encapsulated packet streams (e.g. Ethernet, IP, MPLS).

The traditional Asynchronous Mapping Procedure (AMP) supported in today's OTU2 and OTU3 line ports is complemented with the new GMP process, enabling support of ODU0, ODU2e and ODUflex signals by those line ports.

GMP is also the mapping procedure used to map new sub-2.5G CBR clients (e.g. 1GE, STM-1/4, FC-100/200, CPRI Opt. 1/2/3, ..) into ODU0/ODU1, and to map 40GE and 100GE into ODU3 and ODU4.

Other new CBR clients (e.g. FC-400/800, CPRI Opt. 4/5/6, 3G SDI, ...) are mapped into an ODUflex using the very simple Bitsynchronous Mapping Procedure (BMP), which wraps the ODUflex around the CBR client.

ODUk and ODUflex will also used to carry Ethernet VLANs, MPLS(-TP) LSPs and PWs and IP/IPVPNs via GFP-F encapsulation between L2 PE and L3 PE switching fabrics in either L2 or L3 PE nodes, or embedded in the OTN XC nodes. Those ODUk/ODUflex connections act as tunnels (OTN (sub) Lambda Switched Paths (OTN sLSP/LSP) with GMPLS control) for those packet service layer signals. Use of ODUflex connections between two L2/L3 PE fabrics allows to resize those sLSP tunnels when bandwidth demand grows.

OTN (s)LSP type tunnels are economical when tunnel bandwidth is larger than 0.4G. Below 0.4G MPLS(-TP) Label Switched Path (MPLS LSP) type tunnels are more economical. OTN sLSP and OTN LSP connections are both supported by the LO ODU layer in the OTN. L2 and L3 PE node ports with OTN sLSP/LSP endpoints are connected to the OTN via OTUk (k=2,3,4) Inter Domain Interfaces (IrDI) specified in G.709V3. The OTU3/4 IrDIs use the same multi-lane interfaces as 40GE/100GE.

The ODUflex and its enabling GMP technology are at the core of the next gen OTN technology, and will be used as basis for any future circuit based transport network. With those technologies circuit networks combine the traditional high performance connections with pseudowire like bandwidth flexibility.

ODUflex/GMP are technologies used in the core of the OTN enabling support for any service rate. VCAT/LCAS are technologies used at the edge of the SDH network to support service rates that the core did not support.

ODUflex will as such be very different from VCAT/LCAS.


HSalemi 12/5/2012 | 4:44:20 PM
re: AlcaLu Joins the War for the Optical Core How long do you think it will take to deploy ODUFlex in real networks?

The experience of VCAT/LCAS has been very long and if I am not mistaken, doesn't have a huge usage within sonet/sdh networks.

Why ODUFlex will be different?
mvissers 12/5/2012 | 4:44:21 PM
re: AlcaLu Joins the War for the Optical Core

The OTN switch will get or already has L2 switching capability to support EVPL, EVPLAN, EVPT, EPT and EPLAN services at a minimum.

Routers have Lambda type interface ports on it, to integrate with the initial WDM capabilities in the core. To integrate the router with the emerging switched OTN, routers will be equipped with sub Lambda interface ports, which provide the Gbit/s LSP tunnels between pairs of service routers. Lambda type ports will be used with the emerging switched OTN when two service routers are to be interconnected by 10G, 40G or 100G tunnels.


hyperunner 12/5/2012 | 4:44:22 PM
re: AlcaLu Joins the War for the Optical Core Pera - DWDM in routers is a bad idea. Always better to keep services separate from transport.

Cisco will keep pushing it, and everyone but Sprint will keep ignoring them.

pera 12/5/2012 | 4:44:22 PM
re: AlcaLu Joins the War for the Optical Core

The putting of packet capability to the optical switch or put DWDM to the router?

mvissers 12/5/2012 | 4:44:23 PM
re: AlcaLu Joins the War for the Optical Core

In todays packet core networks MPLS LSPs are used to connect service routers around the core with each other. Those LSPs provide virtual ports within the 10G, 40G and 100G physical interface ports.

With the ever growing bandwidth of these LSPs in the packet core networks it is becoming more efficient to migrate those LSPs to sub-Lambda Switched Paths (sLSPs). The 10G, 40G and 100G physical interface ports will then be equipped with sLSP based virtual ports.

The technology underlying those sLSPs is the ODUflex technology. These ODUflex connections start/end then in the service router. An example is described in http://www.huawei.com/broadban....

With the 40G and 100G physical interface modules for ethernet and OTN being the same devices, the change of virtual port technology does not change the principle architecture of the packet core from the viewpoint of the service router (LSPs are complemented and/or replaced by sLSPs, control plane is GMPLS). From the optical core perspective there is now no need to go to the packet level, keeping all processing on the layer 1 level.


sdmitriev 12/5/2012 | 4:44:23 PM
re: AlcaLu Joins the War for the Optical Core

Alcatel-Lucent has a White Paper: "Overcoming the cost-capacity crunch Implementing the intelligent Optical Transport Network (OTN) to face the "exaflood" phenomenon and support sustainable IP traffic growth", which describes the way 1870 will handle all different kinds of traffic. Take a look  here.

photon2 12/5/2012 | 4:44:23 PM
re: AlcaLu Joins the War for the Optical Core

If the 1870 switches packets like the 1850 did, it will be a disaster.  Hopes are they upgraded the switch fabric enough (not just the chip) to handle it.  My sources tell me the 1850 fell over with only slight increases in packet traffic.  Bet Verizon will put it to a hard test soon, and we'll all find out if this works or not....


Alberto Valsecchi 12/5/2012 | 4:44:23 PM
re: AlcaLu Joins the War for the Optical Core 1870 TTS as announced today is focusing on providing 4Tbps electrical switching capability and is leveraging 1830 Photonic Service Switch (PSS) and other Alcatel-Lucent's DWDM platforms to achieve integrated OTN photonic-electrical switching. Further integration is planned and will follow the market demand.
Tunable XFPs make the integration with DWDM available today
jepovic 12/5/2012 | 4:44:24 PM
re: AlcaLu Joins the War for the Optical Core

Isn't one of the most obvious lessons learned from the history of telecom that god boxes suck?

And what's the point of almost being able to replace my level 2 network?

Let's hope the level 1 fucntionality kicks ass.

Sterling Perrin 12/5/2012 | 4:44:24 PM
re: AlcaLu Joins the War for the Optical Core


The 1870 and the 1850 are both built with a universal/agnostic switching fabric that can do any mix of TDM switching and Ethernet/packet switching, so the abilities are definitely beyond Ethernet over SONET/SDH or over OTN. I agree that products that simply map Ethernet to TDM are not true packet-optical boxes.

Separately, on Andrew's point, my understanding is that the 1870 will be used initially with the 1830 to do DWDM transport, so, initially, the 1870 is a switch only.

It would be helpful if someone from ALU could chime in to clarify DWDM transport on the 1870. Any takers?


torivar 12/5/2012 | 4:44:24 PM
re: AlcaLu Joins the War for the Optical Core <div>Is this a larger version of the TSS product from ALU? &nbsp;It seems to marry the same sets of technologies in a much higher capacity package. &nbsp;</div>
<div>It still needs to employ some sort of L2 or L3 circuit behavior using 802.1Qay or MPLS-TP/regular MPLS in order to statmux and keep the network from being just a big L2 switch...&nbsp;</div>
chechaco 12/5/2012 | 4:44:24 PM
re: AlcaLu Joins the War for the Optical Core

Glad we're in agreement. In regard to use of the agnostic fabric (it's actually packet fabric though very fast) I'd slightly modify "can do any mix" into "potentially can do any mix". I'm sceptical because of ability to support TDM services over the agnostic (packet) fabric in a scalable and cost effective way.


chechaco 12/5/2012 | 4:44:25 PM
re: AlcaLu Joins the War for the Optical Core

I believe that for the real P-OTS, core or metro, just "handling Ethernet" is not good enough. What could it mean? Variety of Ethernet UNIs? Mapping Ethernet services onto SONET/SDH or ODUx tunnels? But that is not real Packet Transport. Smoke and Mirrors.

Stevery 12/5/2012 | 4:44:25 PM
re: AlcaLu Joins the War for the Optical Core

They're at the front of the whole, for lack of a better term, God Box that combines the Layer 2 and some of the Layer 0 functionality,"

I confess:&nbsp; I have no idea how you combine photon/electron level transport with something that involves a MAC address.&nbsp; I would have said the traditional thing:&nbsp; I use layer 0 (really layer 1) to implement my layer 2.

Clues welcomed.

Mark Sebastyn 12/5/2012 | 4:44:25 PM
re: AlcaLu Joins the War for the Optical Core

The system definitely supports DWDM. It does this through the use of pluggable XFP's. While this certainly can't match the performance of high-end 300-pin tunable based solutions, I'd say it does qualify it as DWDM capable.

Sign In