1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Broadcom Unveils World's Highest Density 100GbE Switch Solution

Discussion in 'News' started by Cristian_25H, Apr 30, 2012.

  1. Cristian_25H

    Cristian_25H News Poster

    Joined:
    Dec 6, 2011
    Messages:
    4,626 (4.25/day)
    Thanks Received:
    1,164
    Location:
    Still on the East Side
    Broadcom Corporation, a global innovation leader in semiconductor solutions for wired and wireless communications, today introduced the BCM88650 series, the world's highest density 100 gigabit Ethernet (GbE) switching solution, enabling the design of switching platforms with densities up to 4,000 100 GbE ports. With the industry's highest level of integration, the BCM88650 system on chip (SoC) combines the features and functionality of a complete line card into a single chip. Together with Broadcom's leading FE1600 (BCM88750) fabric, the BCM88650 SoC enables a new generation of high density networking solutions exceeding 100 terabits per second (Tbps). See the technology in action at INTEROP 2012 or visit http://go.broadcom.com/interop12 to learn more.

    [​IMG]

    As the popularity of social networking, streaming video and high bandwidth business services continue to climb, demand for higher-speed networks is growing at an astounding pace. As a result, scalable and affordable 100GbE platforms are a key requirement for next generation switching infrastructures. Large data centers with thousands of servers require 100 Gbps network connectivity in the core to the edge, while large service provider networks require high density core switching platforms with 100 Gbps interfaces to support the increasing access capacities such as 10G PON. The BCM88650 series is the only merchant silicon solution that can process a single stream of 200 Gbps traffic at Layer 2-Layer 4 with integrated advanced packet classification and deep-buffer traffic management features to support data center, carrier Ethernet and packet transport requirements.

    Analysts expect 100 GbE technology to significantly outpace the growth of 40 GbE in its first years of introduction. In many ways, 40G has acted as a trailblazer for 100G, reducing risk at the component level and familiarizing service providers and test equipment vendors with coherent networking.

    Unparalleled Integration Reduces Board Space, Power and System Cost

    The unparalleled integration of the BCM88650 allows OEMs to reduce board space and power while lowering system costs. The unified infrastructure enables system vendors to build a single, scalable product line sharing the same switching infrastructure to address a variety of densities and applications.

    A carrier access switching solution with a total capacity of a few hundred Gbps can be designed with a single BCM88650 device. A chassis with a total capacity of 25 Tbps can be designed for the core of the data center or the carrier core network. Moreover, multiple chassis of different capacities can be interconnected to create a scalable core platform and deliver up to 4,000 wire-speed ports of 100 GbE or their 40 GbE/10 GbE equivalent.

    Market Drivers:

    - Global IP traffic expected to grow 20-fold by 2016
    - 1 million minutes of video content will cross the network every second by 2015
    - The number of devices connected to IP networks will be 2x the global population by 2015
    - Massive traffic growth challenges current network architecture
    - 10/40/100 GbE revenues are expected to reach approximately $50B by 2015

    Key Facts:

    - Integrates advanced packet classification, deep-buffer traffic manager, and cell based fabric interface
    - Integrated 1/10/40/100 GbE network interfaces eliminates the need for additional components
    - Programmable packet classification engine with built-in support for data center, metro Ethernet
    and transport applications
    - Large on-chip classification databases can be extended using a companion Broadcom/NetLogic processor
    - Deep buffers with distributed scheduling scheme allowing state of the art hierarchical QOS, transmission scheduling and flow control schemes
    - Fully compatible with Broadcom's industry-leading XLP multi-core processors and NL8865x knowledge-based processors for best-in-class control plane and expanded L2-4 header processing performance
    - Fully supported by Broadcom's common Application Programming Interface (API)

    Availability:

    The BCM88650 series consists of multiple variations, each tailored for a specific market segment or application including data center, enterprise, carrier access, metro Ethernet, and transport applications. All devices are now sampling with production volume slated for the second half of 2012.
     
    1c3d0g says thanks.
  2. v12dock

    v12dock

    Joined:
    Dec 18, 2008
    Messages:
    1,611 (0.74/day)
    Thanks Received:
    321
    Still waiting on 10gbit nics
     
  3. Aquinus

    Aquinus Resident Wat-man

    Joined:
    Jan 28, 2012
    Messages:
    6,686 (6.46/day)
    Thanks Received:
    2,340
    Location:
    Concord, NH
  4. 1c3d0g

    1c3d0g

    Joined:
    Dec 9, 2007
    Messages:
    699 (0.27/day)
    Thanks Received:
    59
    Odd. I thought Ethernet topped out at 40 Gb/s before other more exotic networking protocols had to be used. Guess they broke through that barrier, which IMO can only be good news. :)
     
  5. Aquinus

    Aquinus Resident Wat-man

    Joined:
    Jan 28, 2012
    Messages:
    6,686 (6.46/day)
    Thanks Received:
    2,340
    Location:
    Concord, NH
    I consider Infiniband exotic. :)
     
  6. Aether New Member

    Joined:
    Jun 1, 2010
    Messages:
    14 (0.01/day)
    Thanks Received:
    3
    I believe, as far as physical media is concerned, the current "limit" (used very loosely) is roughly 25.78Gbps on Single Mode Fiber (or 25Gbps at the MAC). What they do in 40G/100G Ethernet is they use multiple fiber links (either 4 full duplex links or 10) and the Ethernet frames are transmitted in a cascaded/parallel structure. The physical layer reconstructs the parallel data into serial data which is then passed up to the higher OSI layers. It is "essentially" a modified version of 10GBASE-R which adds in some restrictions to 64b/66b and adds "Alignment Markers" to handle de-skewing the parallel links so that the data is received/processed in a coherent manner.

    With 100GbE, you can either use 4 full duplex links using high quality optics and single mode fiber, or 10 full duplex links using cheaper quality optics and multimode fiber, or you can also go electrical and use a big ass cable with a total of 40 conductors not including signal grounds and shielding (10 differential transmitters + 10 differential receivers).
     
    1c3d0g says thanks.
  7. Neuromancer

    Neuromancer

    Joined:
    May 23, 2008
    Messages:
    379 (0.16/day)
    Thanks Received:
    64
    Location:
    South Jersey
    Still waiting on truly affordable 1 GbE to the house.
     
  8. a111087

    a111087

    Joined:
    Apr 2, 2007
    Messages:
    2,766 (0.99/day)
    Thanks Received:
    201
    Location:
    US
    fixed it ;)
     
  9. MxPhenom 216

    MxPhenom 216 Corsair Fanboy

    Joined:
    Aug 31, 2010
    Messages:
    10,132 (6.54/day)
    Thanks Received:
    2,322
    Location:
    Seattle, WA
    wow, I was at microsoft job shadowing for the new job im hoping to get and job shadowed a Network Engineer there and they had 10GBe and all that. 100GbE would be nuts!
     

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)

Share This Page