Architectural Support for 50 GbE and 100 GbE Breakout

Similar documents
Revisi&ng MCBHCB Requirements in Support of 50G/lane PAM4

Alternate Path to Consensus on 400 GbE PMDs

COM Analysis on Backplane and Cu DAC Channels

Updated 50G PAM4 C2M Simulations

Pivotal Issues for 400 Gb/s Ethernet

Impact of Clock Content on the CDR with Propose Resolu8on

400GBase-LR8: A Proposal for 10 km Objective Using 50 Gb/s PAM4 Signaling

Feasibility of 30 db Channel at 50 Gb/s

Optical Trends in the Data Center. Doug Coleman Manager, Technology & Standards Distinguished Associate Corning Optical Communications

Issue with 50G PAM4 C2M Specification

Update of Bit multiplexing in 400GbE PMA

P802.3bs Nomenclature

Path to Consensus on 400 GbE PMDs

400GE at the Cost of 4 x 100GE

Nomenclature Debate. IEEE P802.3bs Task Force. John D Ambrosia Futurewei, a subsidiary of Huawei. IEEE 802 Jul 2016 Plenary San Diego, CA, USA.

Objectives for NG 200G and 400G PHYs

25 Gb/s Ethernet Study Group CSD: Technical Feasibility. Joel Goergen - Cisco Matt Brown APM Version goergen_25ge_01_0914d

Path forward for 100G SMF objectives in 802.3cd

Investigation on Technical Feasibility of FEC Architecture with 1X400Gbps or 4X100Gbps

Ethernet Adoption: SERDES Rates & Form Factors

400G PAM4 The Wave of the Future. Michael G. Furlong - Senior Director, Product Marketing

New Data Center and Transport Interconnect Technology

Transmi(ers Compliance and Receivers Stress Sensi5vity

Timeline Implications

An Alternative Scheme of FOM Bit Mux with Pre-interleave

Path forward for 100G SMF objectives in 802.3cd

FEC performance on multi-part links

802.3bj FEC Overview and Status. 1x400G vs 4x100G FEC Implications DRAFT. IEEE P802.3bs 400 Gb/s Ethernet Task Force. Bill Wilkie Xilinx

Transmitter testing for MMF PMDs

The Ethernet Rate Explosion John D Ambrosia

MPM-400G 400G Multi-Protocol Module

Thoughts for 400GbE FEC Architecture with Pre-interleave

100G-Optics Next Gen Possible PMD Set

Fiber Optic Cabling Systems for High Performance Applications

100 GBE - WHAT'S NEW AND WHAT'S NEXT

Investigation on Technical Feasibility of FEC Architecture with 1X400Gbps or 4X100Gbps

ETHERNET ROADMAP UPDATE. Scott Kipp October 2014

Feasibility of Unretimed 4x25G Host to Module cppi-4

IEEE P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s Ethernet 2nd Task Force review comments

Disruptive Technologies Shaping Data Center Growth

Ethernet Alliance Technology Exploration Forum 2014 The Rate Debate

IEEE Liaison report (Oct ) T v000 Tom Palkert, Macom

Next-gen 400 and 200 Gb/s PHYs over Fewer MMF Pairs Call For Interest Consensus Presentation. IEEE Draft 0.3

Technical Feasibility of optical PMDs with RS FEC

Architecture: Configurations and use cases

Architecture: Configurations and use cases

Arista 400G Transceivers and Cables: Q&A

Ethernet evolution: what s new and what s next

Ethernet The Next Generation

FEC Core Area Comparison and Model

100GEL C2M Channel Analysis Update

Xu Wang Hardware Engineer Facebook, Inc.

802.3bj FEC Overview and Status AMP_Valid and PCS_Lane TBDs DRAFT IEEE P802.3bs 400 Gb/s Ethernet Task Force Mark Gustlin - Xilinx

100Gbps/Lane Electrical Signaling. Loop-Timed Configurations. Ramin Farjadrad SVP of R&D

Consensus building discussion on next-generation MMF PMDs. Presented by Robert Lingle, Jr. IEEE Interim Huntington Beach, CA January 9 th, 2017

Accelerating the Mass Transition of 100 GbE in the Data Center. Siddharth Sheth, Vice President of Marketing, High-Speed Connectivity

OIF CEI-56G Project Activity

IEEE 802.3by 25G Ethernet TF A BASELINE PROPOSAL FOR RS, PCS, AND FEC

100G and Beyond: high-density Ethernet interconnects

Considerations for 25GbE

50G and 100G Use Cases. Brad Booth Azure Networking, Microsoft

100 Gbps/λ vs. 50 Gbps/λ : Cost and Power Comparisons in Silicon Photonics. Brian Welch

Towards building a low cost 25G base PHY for 100G EPON

OTN Support Update P802.3bs 400 Gb/s Ethernet Task Force. Steve Trowbridge Alcatel-Lucent

S2C K7 Prodigy Logic Module Series

Scaling the Compute and High Speed Networking Needs of the Data Center with Silicon Photonics ECOC 2017

Pay as You Grow or Mortgage Model? - The Needs for Standalone 25G EPON

Business Case Components

Single Chip Built In Fet Type Switching Regulator Series

100 GBE AND BEYOND. Diagram courtesy of the CFP MSA Brocade Communications Systems, Inc. v /11/21

802.3bj FEC Overview and Status 400GbE Logic Challenges DRAFT IEEE

Effec%ve Replica Maintenance for Distributed Storage Systems

In Support of a 500m objective for 200G Ethernet: Part II

The zettabyte era is here. Is your datacenter ready? Move to 25GbE/50GbE with confidence

Edwin Malle+e, Bright House Networks.

128GFC. Fiber Channel Leading Charge for Higher Speeds Skip Jones Date: Tuesday, October 15 Time: 2:10pm 2:50pm Location: Room 102B

Integrated Optical Devices

Next-gen 200 and 400 Gb/s PHYs over Fewer MMF Pairs Call For Interest Consensus Presentation

The equivalent of XAUI and other architectural considerations in 40G/100G Ethernet

Roadmap to 400 Gigabit Ethernet over Multimode Fiber WHITEPAPER

Cubro Packetmaster EX48600

25Gb Ethernet. Accelerated Network Performance and Lower Costs for Enterprise Data Center and Cloud Environments. White Paper

Parallel Real-Time Systems for Latency-Cri6cal Applica6ons

QSFP-DD Test Development Kit

Ini$al Deployment and Growth Scenarios for EPoC

Latency and FEC options for 25G Ethernet

Network Coding: Theory and Applica7ons

What were his cri+cisms? Classical Methodologies:

Decoupling electrical and Optical Modulation Andre Szczepanek. Szczepanek_3bs_01a_0315.pdf Decoupling electrical and Optical Modulation 1

Cisco Digital CFP2 Pluggable Optical Module

IEEE Gb/s, 100 Gb/s, and 200 Gb/s Ethernet Task Force

A Distributed Data- Parallel Execu3on Framework in the Kepler Scien3fic Workflow System

16GFC Sets The Pace For Storage Networks

Doug Coleman Manager, Technology & Standards Distinguished Associate Corning Optical Communications

IEEE 802.3ap Backplane Ethernet Overview

IEEE P802.3bs 200 Gb/s and 400 Gb/s Ethernet to RevCom (conditional)

Agenda and General Information

Arista 7500 Series Interface Flexibility

Category 8 paves the way for IEEE 40GBASE-T. David Hughes RCDD/DCDC/NTS Senior Technical Manager MEA CommScope

QDR Active Optical Cables. SC 10 New Orleans

Transcription:

Architectural Support for 50 and 100 Breakout Ali Ghiasi Ghiasi Quantum LLC IEEE 802.3cd Task Force Mee@ng San Diego July 26, 2016

Overview q Baseline for 50 and NG 100 adopted in Whistler h=p://www.ieee802.org/3/cd/public/may16/nicholl_3cd_01a_0516.pdf Leveraging CL82 40 for 50 having 4 lanes Leveraging CL82 100 for NG 100 having 20 lanes Adopted s are compaqble with P802.3bs architecture and supports both 25G and 50G IO q In Whistler we didn t make a decision on # lanes or distribu@on to further study the burst error not using the P802.3bs interleaved TX differenqal pre-coding was invesqgated for 50 Gb/s link h=p://www.ieee802.org/3/cd/public/adhoc/archive/hegde_070616_3cd_01_adhoc.pdf RS(544,514) performance with pre-coding and bit/symbol were invesqgated h=p://www.ieee802.org/3/cd/public/adhoc/archive/anslow_070616_3cd_01_adhoc.pdf TX differenqal pre-coding helps burst error on links with large dominant 1 st tap allowing to build a supporqng bit-mux and compaqble with P802.3bs A supporqng bit mux is key to building common ports/modules and breakout applicaqons q In this contribu@on the need to support 25G lanes in support of breakout applica@ons will be discussed. A. Ghiasi 2 IEEE 802.3cd Task Force

SupporQng Breakout ApplicaQons q QSFP56 4x50G (200 ) 2x100 4x50 q QSFP-DD 8x25G (200 ) 2x100 4x50 8x25 (in the Qme frame not likely applicaqon) q CFP8 16x25G (400 ) 2x200 4x100 8x50 16x25 (in the Qme frame not likely applicaqon) q Next will examine architectural requirement to support breakout. A. Ghiasi 3 IEEE 802.3cd Task Force

SupporQng Breakout in CFP8 q baseline already supports 25G lane h=p://www.ieee802.org/3/cd/public/may16/nicholl_3cd_01a_0516.pdf q also needs to support 25G lane Currently 200 /400 support 25G lanes and allow implemenqng into 25G IO host ASIC Unless 50 and NG 100 supports 25G lane then is forced into the CFP8 module resulqng in more complex module architecture recommended to support 25G lanes in support of breakout applicaqon, flexibility, compaqbility with BS, and allow to reside on host ASIC or in the - chip. Super-set Host Super-set SOC + 400 2x200 4x100 8x50 CDAUI-16 2xCCAUI-8 4xCAUI-4 8x50GAUI-2 - Mux + (not desirable) CFP8 CDAUI-8 2xCCAUI-4 4xCAUI-2 8x50GAUI-1 400 2x200 4x100 8x50 A. Ghiasi IEEE 802.3cd Task Force 4

SupporQng Breakout in QSFP-DD q baseline already supports 25G lane h=p://www.ieee802.org/3/cd/public/may16/nicholl_3cd_01a_0516.pdf q also needs to support 25G lane Currently 200 /400 support 25G lanes allow implemenqng into 25G host ASIC Unless 50 and NG 100 supports 25G lanes then is forced into the QSFP-DD module resulqng in more complex module and possibly exceeding QSFP-DD power envelope architecture recommended to support 25G lanes in support of breakout applicaqon, flexibility, compaqbility with BS, and allow to reside on host ASIC or in the - chip. Super-set Host CCAUI-8 2xCAUI-4 4x50GAUI-2 QSFP-DD Super-set SOC + 200 2x100 4x50 - Mux + (not desirable) CCAUI-4 2xCAUI-2 4x50GAUI-1 200 2x100 4x50 A. Ghiasi IEEE 802.3cd Task Force 5

SupporQng Breakout in QSFP56 q baseline already supports 25G lane h=p://www.ieee802.org/3/cd/public/may16/nicholl_3cd_01a_0516.pdf q QSFP56 having na@ve 50G IO is designed to operate with na@ve 50G host IO but early ASIC implementa@on my require 2:1 - mux Currently 200 /400 support 25G lanes and allow implemenqng into 25G host ASIC Any 2:1 - mux/ on the line card it will not add complexity or power to the QSFP56 architecture supporqng 25G lanes provide benefit even in this scenario by allowing opqmizing the implementaqon based on availability host ASIC/FPGA and - Mux. Super-set Host Super-set SOC + 200 2x100 4x50 CCAUI-4 2xCAUI-2 4x50GAUI 2:1 PAM- Mux + (not desirable) IC QSFP56 200 2x100 4x50 A. Ghiasi IEEE 802.3cd Task Force 6

SupporQng Breakout q In addi@on to the suppor@ng 25G lanes for ease of implementa@on needs to support 25G lanes as well! Super-set Host with 25G IO Module Common Suppor@ng 50, NG 100, 200, and 400 25G Lanes Supported 8x50 4x100 2x200 400 8x50 4x100 2x200 400 25G Lanes Not Supported 8x50 4x100 2x200 400 8x50 4x100 2x200 400 A. Ghiasi IEEE 802.3cd Task Force 7

Summary q Commonality of and where possible with 802.3bs allow broadest set of breakout applica@ons 50 will be most common breakout from 200G ports 100 will be the most common breakout from 400G ports q Adhoc inves@ga@on of (544,514) with differen@al pre-coding indicate bit mux can be support on non-interleaved 50 /NG 100 allowing to build common /module suppor@ng P802.3bs applica@ons q The 50 and NG 100 supports 25 lane which allow early implementa@on as well as more efficient 2 nd genera@on implementa@ons based on 50G IO q Recommend to also support 25 lane for compa@bility with P802.3bs and allow breakout applica@on using a common IC without the need to force into the module. A. Ghiasi 8 IEEE 802.3cd Task Force