SDN architectures

Ethernet Europe: Top 10 Takeaways

FRANKFURT -- Ethernet Europe 2013 -- The advance program for this year's Ethernet Europe, which has attracted more than 300 attendees to Marriott Frankfurt Hotel on Day 1, made it quite clear that software-defined networking (SDN) was going to be discussed and debated at the event just as much as the more traditional Carrier Ethernet topics.

So here are the 10 key issues that captured my attention during the conference sessions, and on the show floor, during the first day's proceedings.

  • Get ready for Bandwidth-on-Demand: It's clear from a number of presentations, including those from TeliaSonera International Carrier (TIC) and the event's host, Heavy Reading senior analyst Stan Hubbard (who has many insights into service provider plans), that dynamic bandwidth provisioning is going to feature strongly in carrier portfolios very soon. While such services already exist, they have tended to be based on technology from a single supplier: The difference now, according to presenters and attendees here, is that the industry is closing in on a common set of tools for the provision of bandwidth-on-demand services in multi-vendor networks.

  • The MEF gets cloudy: The Metro Ethernet Forum (MEF) is preparing an "earth-shattering" announcement regarding Carrier Ethernet 2.0 and cloud services in the near future. This is likely to be very closely related to the bandwidth-on-demand trend, as service providers figure how they can provide flexible Ethernet services that can be turned up and down to match the way in which their enterprise users use cloud-based resources and applications.

  • SDN isn't a BGP killer: Nicolas "Nico" Fischbach, director of Network Platform & Strategy and Architecture at Colt Technology Services Group Ltd., put some minds at rest about the potential impact of SDN. "People have been saying there will be no BGP [border gateway protocol] and that IP networks as we know them will go away. That is NUTS!" he proclaimed. "BGP will never go away. But it may be combined with something else, possible OpenFlow but there are lots of options," added the Colt man. Fischbach is a man worth listening to, as Colt has been checking out SDN use cases to figure out any potential advantages and is very active in the network functions virtualization (NFV) Industry Specifications Group formed in late 2012 under the auspices of European Telecommunications Standards Institute (ETSI). (See Carriers Peer Into Virtual World.)

  • Does OpenFlow have a shelf life? Fischbach soon got into his stride once he was up on stage, questioning whether SDN's poster-child protocol OpenFlow will be around for long. "I'm not sure about the future of OpenFlow as a protocol. I'm not sure it's the SDN Holy Grail." Plenty of speakers, including Rafael Francis from Cyan Inc., noted that there are other tools that will be utilized in SDN architectures.

  • Where are the carrier class OpenFlow products? There was no stopping the day's most compelling speaker. Fischbach told conference delegates that while there is plenty of market noise around OpenFlow products, "carrier-class OpenFlow switches are not GA [generally available]." That sounds like an invitation for some vendors to at least try to prove him wrong…

  • Vendors have been distracted by SDN: Fischbach again. Colt's new network architecture includes an integrated packet/optical core, which isn't quite fully deployed yet because the vendor community is lagging behind its timelines by about two years. SDN has "caused chaos… [vendors] have jumped on the bandwagon and change their positioning" to ensure they have a SDN story.

  • Major carriers are serious about NFV: Network functions virtualization (NFV) isn't just something the ETSI group's members are meeting and talking about – they're checking out whether dedicated telecom products with the traditional integrated hardware/software combination can be replaced with a combination of centralized applications, virtualization software and distributed off-the-shelf servers. And according to Karl-Heinz Nenner, senior architect at Deutsche Telekom AG, the carriers are finding that existing server technology is robust enough for deployment in an NFV architecture.

  • SDN + CE 2.0 is the future: Heavy Reading's Stan Hubbard has seen the future of Carrier Ethernet services and, if SDN capabilities are indeed introduced into carrier networks alongside the enhanced management capabilities provided by Carrier Ethernet 2.0 specifications, users will be offered network-as-a-service options within a few years. CE 2.0 specifications are already enhancing performance management capabilities, added Hubbard.

  • Say hello to 'Anywhereization': We're not sure how hot this term, being used extensively by the TeliaSonera International Carrier (TIC) team, will be, but it encapsulates what all the major operators are seeing and having to deal with – that customers are demanding flexible, increasingly high-bandwidth services anywhere, any time and on any device. The days of fixed rate services over point-to-point connections are disappearing…

  • He called himself 'The Dude': Tata Communications Ltd. senior director and head of Ethernet Product Management, Henry Bohannon, has made the Light Reading team very happy. "My favorite movie of all time is The Big Lebowski," he told me Tuesday afternoon. We believe he had a rug and it really tied his room together.

    — Ray Le Maistre, Editor-in-Chief, Light Reading

  • Page 1 / 2   >   >>
    PhilGr 4/17/2013 | 7:03:10 PM
    re: Ethernet Europe: Top 10 Takeaways Sorry - got to call BS on that. Bandwidth on demand services have been available for longer than Infinera has been in business - just look at the history of ATM, RSVP and MPLS-TE. The capabilities have evolved over time. Schlumberger launched a service back in 2004 and I'd guess that wasn't the first. -á -á -á
    Alex_Fduch 4/17/2013 | 2:16:43 PM
    re: Ethernet Europe: Top 10 Takeaways -áOne more point.
    People forgot that in the SW world virtualization became a booming trend in Data Centers only after Intel and AMD added vitrualization support in their CPUs, or in other terms - in Hardware! Before that virtualization was slow and people would never use it for high performance apps, they deployed dedicated HW servers for that - one server for one application.

    So, IMHO, that means NFV will become comparable with existing networking HW in scalability, performance and efficiency when CPU/NICs/Server chip sets vendors will add NPU functions in their products.
    Ray Le Maistre 4/17/2013 | 11:06:20 AM
    re: Ethernet Europe: Top 10 Takeaways Alex - great post.-á

    And this is what the service providers participating in the NFV group are exploring right now -- they want to figure out if the reality can match the perceived potential.-á

    You are totally right to be skeptical and my sense is that the operators are very wary of getting carried along without drilling deep into the potential and use cases.

    There are clearly some applications/functions that are being explored initially, particularly related to mobile packet core functions.
    Deutsche Telekom provided an update on NFV at this week's Ethernet Europe in Frankfurt and I will be reporting on that update shortly.-á
    Alex_Fduch 4/17/2013 | 9:52:57 AM
    re: Ethernet Europe: Top 10 Takeaways Just for those, who wants less hype and more details about issues of NFV, I've found a good discussion:
    Alex_Fduch 4/17/2013 | 9:17:36 AM
    re: Ethernet Europe: Top 10 Takeaways Nonsense from my perspective. If "network functions in software that can run on a range of industry standard server hardware" would be really better and cheaper than HW, we should have already seen SW routers replacing HW routers (BRAS, BNG, NAT, FW, etc.).
    Remember the IT way:
    Proprietary Mainframes -> PCs -> standardization (de facto) of x86 architecture -> servers and Data Centers on x86 architecture -> Virtualization

    What do we see? We see that standard HW was proven to be faster and cheaper than Mainframes for most of applications.

    While in the networking world we've seen a different path:
    Software routers -> software routers and Hw switches -> HW routers and HW switches.
    If e.g. Vyatta is so great then why it replaces HW routers only in a limited number of enterprise applications? Are customers of networking HW stupid?
    Because networking HW provides you: predictable and stable performance within defined scale and defined operations scope and with low power consumption per bit.

    In current operator world if you build a small network - you can use cheap switches and move a lot of "smart" networking functions on Linux machines. But when your networks growth and hits a large scale your Linux BRAS/NAT/FW farm occupies so much space and consumes so much power and scales so poor that you have to buy a HW solution for your networking functions.
    And existing generic server HW is far behind dedicated networking HW for these functions.

    All presentations and paper's I've read about NFV always talking that generic server HW costs cheaper that is why NFV will be cheaper.
    But how did they compare that?
    Lets compare cost per bit:
    If I have a modern router like Cisco ASR9K or ALU 7750 SR or Juniper MX I can do with one NPU the following:
    1. Routing
    2. Switching
    3. BRAS/BNG
    4. ACL
    5. FW (for Juniper MX)
    And any time I'm adding any of this function in the modern HW (NPU based) router - my performance has no degradation and my power consumption per bit stays the same. And modern NPUs can even switch off some cores and decrease power consumption if they are not fully utilized.

    With NFV - each additional function required more CPU power - so I need to get more and more generic server HW for my networking operations.
    Depending of the quality of SW and CPU power and NICs in the server, etc., etc. Which means I need more power, more ToR switches, more load balancers, etc. in the DC to cope with that.

    I do agree, that we possibly can move some of networking function of some small scale in the DC, virtualize them.
    But we cannot replace ISPs network infrastructure with just a dumb switches and DC, at least not in the next years.
    Ray Le Maistre 4/17/2013 | 7:54:12 AM
    re: Ethernet Europe: Top 10 Takeaways The general feeling here at Ethernet Europe is that SDN is going to find its way into carrier networks very soon -- the opinion of Nico at Colt is that 'carrier class' OpenFlow switches are not 'GA' and I think he knew he was stirring the pot... -á
    Ray Le Maistre 4/17/2013 | 7:45:50 AM
    re: Ethernet Europe: Top 10 Takeaways Just to clarify: This wasn't about picking out or identifying any particular systems vendor -- the point made by several speakers at Ethernet Europe was that, currently, the ability to offer bandwidth-on-demand services has required the deployment of a single-vendor solution, compared with a multi-vendor solution (which is preferred).

    If Infinera has enabled this capability for any service provider then all credit to its R&D team. -á-á
    Ray Le Maistre 4/17/2013 | 7:38:48 AM
    re: Ethernet Europe: Top 10 Takeaways Hi Alex
    Here's a guide to what NFV is about

    What's NFV All About?http://www.lightreading.com/bl...
    tojofay 4/16/2013 | 8:33:07 PM
    re: Ethernet Europe: Top 10 Takeaways -á"is going to feature strongly in carrier portfolios very soon" -á
    Only 1 company does it now and clearly Infinera drove and led the industry. Why no mention? I don't work for Infinera.
    Alex_Fduch 4/16/2013 | 7:34:10 PM
    re: Ethernet Europe: Top 10 Takeaways NFV - is a big hype term which is needed to be clarified.
    Does that mean x86 vs NPU? Or something else?
    Page 1 / 2   >   >>
    Sign In