x
Ethernet equipment

TelecomNext: Even More Ethernet

LAS VEGAS -- As previously mentioned, and as noted again below, IPTV has helped "Carrier Ethernet" become a key buzzword at this week's TelecomNext. (See Alcatel Shrinks Access Router.) Here are a few more doses of Ethernet's March Madness from the show:

  • Nortel Networks Ltd. plans to crank the volume on carrier Ethernet this summer with offerings that likely will focus on the Metro Ethernet Routing Switch 8600, formerly called the Passport 8600. If all goes well, Nortel might spread its ideas to some of its optical gear as well.

    The company wants to specify a scheme for carrier Ethernet that it thinks beats others in its simplicity: Rather than add features, Nortel turns some off, says John Hawkins, senior marketing manager for Nortel.

    What's nifty about this method is that it requires no new standards, Hawkins believes. That could be a big relief to many -- including Hawkins, who's a veteran of the IEEE 802.17 standard wars surrounding Resilient Packet Ring Technology. "It's not going to be a big RPR-like standard," he says. "It's not like we need to do anything to the Ethernet header." (See RPR: Deadlock Ahead? and RPR Moves Forward.)

    The idea is to subtract certain things from Ethernet to make it look more carrier-friendly. For example: Rather than allow Ethernet to learn the surrounding network, as it was created to do, Nortel wants to let carriers dictate network knowledge to switches. Carriers would tell a node where to forward traffic, giving it multiple options so carriers can define protection paths.

    What's left out are pieces such as the Spanning Tree protocol, which was added to Ethernet to prevent loops, or the broadcast flooding of a MAC address to all nodes.

    The reason to do all this is to create determinism, the lack of which "is the main bugaboo Ethernet has always had against it in a carrier environment," Hawkins says. "A frame will get to its destination, but you can't tell me how it will get to its destination, and the carriers are bothered by that."

    The challenge for Nortel is to get other vendors to agree with this deterministic concept, which Nortel is calling Provider Backbone Transport. Nortel's already working on that, Hawkins says.

  • Ciena Corp. (NYSE: CIEN) will be adding more Ethernet to the CN 4200 transport and aggregation box, officials say. The idea is to let the box act as an Ethernet aggregator in places where carriers don't want to spend for a separate Ethernet switch. Some European networks could use this kind of backhaul option for IP DSLAMs, for instance, says Vinay Rathore, director of segment marketing for Ciena.

    The upcoming CN 4200 cards will be meant for those kinds of special occasions only and are not a sign that Ciena plans to go mano a mano against switch vendors like Extreme Networks Inc. (Nasdaq: EXTR) and Foundry Networks Inc. (Nasdaq: FDRY) "It's not a full-blown Ethernet switch," Rathore says.

    Ciena didn't formally announce the Ethernet add-ons, and Rathore won't yet specify when they'll be available.

  • Riverstone Networks Inc. (OTC: RSTN.PK) is awfully proud of its Ethernet cards for the 15008 router. Following the trend of simplified Layer 2 gear, the new cards skip things like IP VPN support and strike for just Layer 2 and virtual private LAN services (VPLS). The cards are aimed at customers using the 15008 just for Ethernet aggregation, those that say they don't want all that Layer 3 stuff, says Inbar Lasser-Raab, Riverstone's vice president of marketing.

    It all ties back to the idea of Ethernet aggregation being a key function for IPTV networks. An Ethernet-heavy architecture is what's helped Alcatel (NYSE: ALA; Paris: CGEP:PA) attract IPTV-related business for its 7750 Service Router and 7450 Ethernet Service Switch. (See Alcatel Router Revenues Surge.) Riverstone wants its piece of that pie.

    "They're doing what the 7450 did, only they're doing it with linecards," says Mark Seery, analyst with Ovum RHK Inc. "Everybody's sort of reacting to Alcatel."

    Oh, the company had some other news, too -- something about Lucent or $207 million or something... (See Lucent Wins Riverstone Auction.)

  • The MEF is adding some IEEE access standards to its certification program, the first time the MEF will be certifying standards that don't originate from MEF-defined criteria. The MEF will check gear for compliance with the operations, administration, and maintenance (OAM) portion of IEEE 802.3ah, the standard for Ethernet in the last mile. It's also going to check vendors for compliance with 802.11t, an upcoming test standard for wireless networks. (See MEF Targets Access Tech.) — Craig Matsumoto, Senior Editor, Light Reading

  • Page 1 / 4   >   >>
    Mark Sebastyn 12/5/2012 | 4:00:48 AM
    re: TelecomNext: Even More Ethernet Taking away features addresses security concerns.

    What is missing is QoS. I think IEEE is now working on end-to-end flow control.
    Pete Baldwin 12/5/2012 | 4:00:48 AM
    re: TelecomNext: Even More Ethernet I don't *think* I'm oversimplifying Nortel's case, although I could be wrong ...

    The nitty-gritty details won't be out for a while, but it looks like they'll use point-and-click provisioning to populate the forwarding tables, rather than letting Ethernet nodes swap MAC addresses.

    It sounds like quite an elegant way to create carrier Ethernet. What's missing?
    filipe 12/5/2012 | 4:00:47 AM
    re: TelecomNext: Even More Ethernet how do they fit in
    metroman 12/5/2012 | 4:00:45 AM
    re: TelecomNext: Even More Ethernet They fit in with French customers, lovers of old 3Com technology and anyone with any investment from France Telecom. Other than that, nice GUI, average product, Nan Chen talking to anyone who will listen and not much else.

    Next on the aquisition heap I think.
    tmc1 12/5/2012 | 4:00:42 AM
    re: TelecomNext: Even More Ethernet Craiger,

    Nortel and John Hawkins are speaking in tongues. They are not simplifying ethernet, that is pure marketing nonsense. They are trying to use IEEE to push new features into ethernet. Features that are already matched by higher level protocols such as mpls and vpls. The problem is, Nortel missed the boat on that (as they often do) and they are trying to move everyone to something they can do. The only feature that makes sense here is the first/last mile OAM function.

    The problem for carriers is that carrier class ethernet gear is expensive. Just like carrier class SONET/SDH gear. The SPs and Nortel want cheap (less feature rich) switches to do this. The problem is that cheap switches are not carrier class and the 8600 certainly is not carrier class and probably never will be. This is the current carrier pipe dream.

    First the market actually needs carrier class ethernet gear (of which there is very little actually specifically designed), then they need to figure out how to drive down the costs of this gear. It will never be as cheap as those home LAN units or cheap enterprise units because the mission is completely different and this is the part that no one wants to accept.
    tmc1 12/5/2012 | 4:00:41 AM
    re: TelecomNext: Even More Ethernet Steve,

    Alcatel has the only purpose built box for this today. Rumor has it that Juniper will introduce one at some point. You can certainly run Juniper M320 or Cisco GSR and they are "carrier-class" but very expensive. All three of these companies understand what it means to build products for these markets.

    Nortel knows how to build carrier class products but they are almost completely clueless about data outside of ATM and SONET.

    Riverstone was one of the early ones trying to build this type of product but they had a weak enterprise product to base it off of. They have never really proven they can build a carrier class product. Now with Lucent they have joined a company that will probably destroy any chance they have.

    The rest of the products are pretty much enterprise plays and L3 switches trying to go where they see all of the hype and money is (IPTV and edge services) but they just don't get it.

    Like I said, I think it keeps coming back to SPs wanting "carrier-class" but they don't want to pay for it. Five nines reliability does not come cheap. Cheap L2 switches with PBT and MAC in MAC will not give you 99.999% uptime... much more is required.
    sgamble 12/5/2012 | 4:00:41 AM
    re: TelecomNext: Even More Ethernet Which vendors would you consider "Ethernet Carrier Class" that are not?

    Alcatel looks strong here especially with their history in the carrier space. Anyone have any details in terms of performance, costs (cap/op) and support? On paper they have a lot of things I am looking for - SLA measurements, H-VPLS support, strong knowledge of IPTV and therefore Multicast issues for a large Metro and so on.

    Foundry is pushing their XMR/MLX and teamed up with Metasolv to show they are in a carrier class position with a strong OSS player for IP Activation, work flow and inventory tracking. But it's Foundry who is an Enterprise Ethernet company by their own definition. MLX was built to compete with Alcatel's Edge version of the Timetra box. They released the XMR without any SAV (Q-in-Q) support (oops!). Their support and QA has not been good. Although I hear they are trying to resolve this as they try to push out of the Enterprise space.

    Nortel is still in a scary space with all their financial and movement of their Management and RD. How many routers (and jobs) have been shuffled in an out of R&D at Nortel over the years?! Neptune, Versalar (spelling) and there was another huge Core router they tried (do not recall the name)?

    Cisco is expensive both in terms of CAPEX and OPEX - especially for an MEU in the Metro. When I have over 400 school sites plugged in at Gigabit Ethernet, the costs to aggregate these on a Cisco box is too much. However, with my limited experience with them, I would say Cisco's support versus any other vendor I have dealt with is fantastic (but man do you pay for it!).

    Juniper has not made a strong move here despite everyone hoping for an acquisition of Foundry (Extreme? ick).

    F10 will be acquired eventually and I can't believe they will IPO (hype to get purchased?). Great port density and cost but will they be around in a year?

    Atrica - read F10.

    There are Mac-in-Mac options now on the pure Ethernet side but Ethernet using RSTP does not scale well in a large Metro with 10+ nodes. The QoS/CoS is weak. Clients connecting with Ethernet switches of their own can cause havoc on the CPU/Core despite vendor attempts to patch these with CPU protection, flooding and so on. Adding these features always seems to create other issues (multicast!).

    VPLS is rolling along nicely and good support from Riverstone and Alcatel to name a couple. But the industry has not agreed on a good solution for Multicast for IPTV from my reading. Even with H-VPLS adding an MTU layer there doesn't seem to be agreement on how the carrier/metro provider should interact with Multicast trees/replication.

    Running a 10Gig Metro Ethernet Core without MPLS was a good start for a greenfield network. But with IPTV and VOIP rolling in the simple 'dumb' 10 Gig pipes are not enough (not with the amount of sites, vlans, traffic patterns, loops and so on that can occur in a switched network). I understand some of the latest generation of switches/cards addresses some of these issues but then I have no room to move to an MPLS network if I need to with some of these.

    But which vendor has "Carrier class Ethernet" in the metro that supports 10Gig at a good price per port and has the best features to protect the switch/router at this point?

    What exactly is the definition of carrier class ethernet? Hitless upgrades? redundant fabrics? non-stop forwarding?

    I have been praying for an LR bake-off again like you they did in the day with Cisco vs. Juniper... Last one they tried only Alcatel showed up and that was a long time ago.

    Do you really need MPLS at the aggregation point? Nortel doesn't seem to think so. That's a lot of LSPs. Most vendors think you do. In the MTU? (or just strong multicast features, Q-in-Q, mac-in-mac).

    So these are my thoughts based on a lot of speculation and reading. What are the opinions of others in this space who know and are moving towards networks supporting IPTV and VOIP?


    Steve.
    sgamble 12/5/2012 | 4:00:41 AM
    re: TelecomNext: Even More Ethernet tmc1,

    First off thanks for the response. It's bang on my thinking and reading to this point.

    The cost of a 10Gig line card from Juniper and Cisco are insane vs. the other vendors I mentioned. You are right - we all want 5x9s but cannot justify the capital expense of upgrading to these when you do not have the prockets of an RBOC.

    My question is how do Alcatel, Cisco and Juniper address broadcast storms and loops initiated by the enterprise clients vs. traditional l2 Enterprise class switches? Add Multicast traffic to the mix and it is a nightmare to deal with from an Operations view. Without the separation of control/data planes this isn't possible it seems.

    I mentioned one scenario I have where 400 schools ALL running l2 switches have the ability today to bring down the entire Core by hijacking CPUs when they have a loop. Or initiate a backup/restore using multicast ghosting to 150 sites over Gig-E. The answer by the vendors in the Enterprise space is to enable CPU protection by enabling features like broadcast/multicast flooding so that the multicast traffic does not get done in software but in hardware. This in turn creates other problems (TCAM DoS and valid dropped multicast frames for people running OSPF as an example when the CPU drops certain frames without discrimination).

    This is the problem with these Enterprise Switch vendors bolting on IP/MPLS from my testing. They are still prone to these issues. Are Alcatel, Cisco and Juniper any different? So if I have 400 schools plugged into 7450s from Alcatel running a mesh of LSPs how would this evil scenario be any different... Would the head-end 7450 where the backup is initiated from suffer the same ill affects but the rest of the network would be fine?

    From my reading it's not only my vendor that has this issue. Topology has a lot to do with it as well. Looking at the H-VPLS drafts and vendor white-papers I am still curious based on the scenario I gave above how these "Carrier Class" Ethernet vendors are any better.

    Sure I understand the base VPLS/VLLs will give me in terms of QoS, Fast-Reroute, Hitless upgrade, and VLAN numbering. But having a large Enterprise customer be able to knock out the network is not Carrier Class. This is my biggest issue I need to resolve and why I am looking into Alcatel, Cisco and Juniper. And with Multicast applications ramping up over the last year and continuing this way I am concerned when the industry has not agreed on a solution on Multicast in a l2vpn. Is it too early to invest in an H-VPLS solution in the both the access and core when there are some big scaling issues that are not worked out?

    But the literature on MPLS is mostly marketing literature with little information on how it protects the node from 1 client. If anyone has any white papers or experience from the big three vendors mentioned please post them here...


    Steve.






    OldPOTS 12/5/2012 | 4:00:41 AM
    re: TelecomNext: Even More Ethernet Several other points to consider for 'carrier class';

    )What capabilities do they have for a large volume (carrier levels) of subscribers?

    )How easily can a customer and his services (streams) be configured? (Remember carrier levels)

    )How well do they support different streams to a customer? (Service QoS)

    )How easily can OAMP be performed? (Again carrier leve) (May require 3rd party SW)

    )How well does it manage alarms to monitoring system? (not overwhelming)

    This usually has to be demoed in a 'carrier' configuration that is larger than a few nodes.

    Also Video requires more than 5-nines to compete with cable.

    OP
    sgamble 12/5/2012 | 4:00:40 AM
    re: TelecomNext: Even More Ethernet )How easily can a customer and his services (streams) be configured? (Remember carrier levels)

    But what is carrier levels in terms of provisioning of services and dealing with MACs?

    )How well do they support different streams to a customer? (Service QoS)

    Couldn't agree more here. Isn't that why MPLS is neeed out to edge vs something like the 8600 running 802.1p?

    )How easily can OAMP be performed? (Again carrier leve) (May require 3rd party SW)

    This is the biggy. Whatever I upgrade to as to help my NOC in troubleshooting issues without an immediate need for a truck roll! This is where my current network fails me.
    Page 1 / 4   >   >>
    HOME
    Sign In
    SEARCH
    CLOSE
    MORE
    CLOSE