Data Center Routing Challenges - LinkedIn

Slides:



Advertisements
Similar presentations
Path Splicing with Network Slicing
Advertisements

Path Splicing with Network Slicing Nick Feamster Murtaza Motiwala Santosh Vempala.
G : DCM Signaling Mechanism Using GMPLS RSVP-TE ITU-T Workshop on IP-Optical, Chitose, Japan 7/11/2002 Dimitrios Pendarakis, Tellium, Inc. ITU-T.
Shortest Path Bridging IEEE 802
CHIARO CONFIDENTIAL Chiaro Networks Dr. Eyal Shekel March 2003.
Data Center Fabrics Lecture 12 Aditya Akella.
Logically Centralized Control Class 2. Types of Networks ISP Networks – Entity only owns the switches – Throughput: 100GB-10TB – Heterogeneous devices:
Deployment of MPLS VPN in Large ISP Networks
Bringing Together Linux-based Switches and Neutron
PortLand: A Scalable Fault-Tolerant Layer 2 Data Center Network Fabric
L3 + VXLAN Made Practical
Introduction into VXLAN Russian IPv6 day June 6 th, 2012 Frank Laforsch Systems Engineer, EMEA
© 2010 Cisco and/or its affiliates. All rights reserved. 1 Segment Routing Clarence Filsfils – Distinguished Engineer Christian Martin –
Data Center Fabrics. Forwarding Today Layer 3 approach: – Assign IP addresses to hosts hierarchically based on their directly connected switch. – Use.
Application Centric Infrastructure
What’s New in BMC ProactiveNet 9.5?
Alan Shieh Cornell University Srikanth Kandula Microsoft Research Emin Gün Sirer Cornell University Sidecar: Building Programmable Datacenter Networks.
Reconfigurable Network Topologies at Rack Scale
Spring 2002CS 4611 Router Construction Outline Switched Fabrics IP Routers Tag Switching.
Interdomain Routing and The Border Gateway Protocol (BGP) Courtesy of Timothy G. Griffin Intel Research, Cambridge UK
A Scalable, Commodity Data Center Network Architecture
Microsoft Virtual Academy Module 4 Creating and Configuring Virtual Machine Networks.
IPv6 Deployment Plan The Global IPv6 Summit 2001.
Chapter 10 Switching and Routing
LIGHTNESS Introduction 10th Oct, 2012 Low latency and hIGH Throughput dynamic NEtwork infrastructureS for high performance datacentre interconnectS.
Keeping Things Simple Is Harder Than You Think Brad Hannah – Manager, Systems and Storage ITServices - Queen’s University April 28 th 2014
Pack-it Technology Highly versatile Internet connectivity solution for non-Internet enabled devices. Seamless interface with RS232, Ethernet, switches,
LAN Switching and Wireless – Chapter 1 Vilina Hutter, Instructor
OIF NNI: The Roadmap to Non- Disruptive Control Plane Interoperability Dimitrios Pendarakis
Evolving Toward a Self-Managing Network Jennifer Rexford Princeton University
Copyright © 2015 Juniper Networks, Inc. 1 QFX5100 Line of Switches The World’s Most Nimble 10/40GbE Data Center Access Switches Speaker Name Title.
Evolving Toward a Self-Managing Network Jennifer Rexford Princeton University
Virtual-Channel Flow Control William J. Dally
Spring 2000CS 4611 Router Construction Outline Switched Fabrics IP Routers Extensible (Active) Routers.
Router Startup and Setup Honolulu Community College Cisco Academy Training Center Semester 2 Version 2.1.
ROM. ROM functionalities. ROM boards has to provide data format conversion. – Event fragments, from the FE electronics, enter the ROM as serial data stream;
Fabric: A Retrospective on Evolving SDN Presented by: Tarek Elgamal.
2014 Redefining the Data Center: White-Box Networking Jennifer Casella October 9, 2014 #GHC
SDN and Beyond Ghufran Baig Mubashir Adnan Qureshi.
Software Defined Networking BY RAVI NAMBOORI. Overview  Origins of SDN.  What is SDN ?  Original Definition of SDN.  What = Why We need SDN ?  Conclusion.
XFabric: a Reconfigurable In-Rack Network for Rack-Scale Computers Sergey Legtchenko, Nicholas Chen, Daniel Cletheroe, Antony Rowstron, Hugh Williams,
Open Networking and SDN
Credits Authors of “Jupiter Rising” [SIGCOMM 2015]:
BGP-Based SPF IETF 97, Seoul
Intelligent Fabric (iFab)
BGP-Based SPF RTGWG - Jan 2017
Segment Routing (SR) Introduction and Tutorial
Hierarchical Fabric Designs
Sebastian Solbach Consulting Member of Technical Staff
Chapter 6 Delivery & Forwarding of IP Packets
6.829 Lecture 13: Software Defined Networking
Network Packet Brokers
DCI using TRILL Kingston Smiler, Mohammed Umair, Shaji Ravindranathan,
The NPD Group - Enterprise DC Agenda
Trellis Tutorial An Open-Source, White-Box, SDN Based Leaf-Spine Network Fabric Charles Chan Open Networking Foundation.
IS-IS Spine-Leaf DC RTGWG Interim
NTHU CS5421 Cloud Computing
Using MPLS/VPN for Policy Routing
The Evolution of the Data Center
GGF15 – Grids and Network Virtualization
MPLS Traffic Engineering
CS 31006: Computer Networks – The Routers
인터넷 구조 2002년 2학기 장주욱.
Link State on Data Center Fabrics
is-is support for openfabric
Separating Routing Planes using Segment Routing draft-gulkohegde-spring-separating-routing-planes-using-sr-00 IETF 98 – Chicago, USA Shraddha Hegde
Use of BIER Entropy for Data Center CLOS Networks
Internet and Web Simple client-server model
RDMA over Commodity Ethernet at Scale
IS-IS VPLS for Data Center Network draft-xu-l2vpn-vpls-isis-02
Presentation transcript:

Data Center Routing Challenges - LinkedIn Russ White Shawn Zandi

Single SKU Data Center Fabric 1 Fabric 2 Fabric 3 Fabric 4 Pod W Spine Leaf Pod X Pod Y Pod Z Single SKU Data Center Fabric 1 Spine Spine Spine Fabric 2 Spine Spine Spine Spine Spine Spine Fabric 3 Fabric 4 Spine Spine Spine 4,096 x100G ports Non-Blocking Scale-out

Complexity within Chassis Chassis: Robust-yet-Fragile Complex due to NSR, ISSU, feature-sets, etc. Larger fault domain, Failover/Fail-back Indeterministic boot up process and long upgrade procedures Moved complexity from big boxes to pizza boxes, where we can easily manage and control! Better control and visibility to internals by removing black-box abstraction! Same Switch SKU on ToR, Leaf and Spine (Entire DC) Single chipset uniform IO design (same bandwidth, latency and buffering) True 5-Staged Clos Topology! with deterministic latency Dedicated control plane, OAM and CPU for each ASIC

Control Plane Complexity at Scale W X Y Z W X Y Z W X Y Z W X Y Z 2304 2305 2306 2307 2336 2337 2338 2339 2368 2369 2370 2371 2400 2401 2402 2403 Pod 1 Pod 11 Pod 21 Pod 31 2048 2049 2050 2051 2088 2089 2090 2091 2128 2129 2130 2131 2168 2169 2170 2171 1 2 … 32 321 322 … 352 641 642 … 672 961 962 … 992

Control Plane Requirements Fast, simple distributed control plane No tags, bells, or whistles (no hacks, no policy) Auto discover neighbors and build RIB Minimal (to zero) configuration Must use TLVs for future, backward compatible, extensibility Must carry MPLS labels (per node/interface)

Control Plane BGP IS-IS Build New Heavy weight; lots of features and “stuff” that are not needed Modifications to support single IP configuration required Does not supply full topology view Proven scaling BGP Not proven to scale in this environment Light weight Most requirements for zero configuration are already met Provides full topology view IS-IS A lot of work But could use bits and pieces from other places Build New

Forwarding Challenges ECMP is blind End to end path selection is required for some applications. Application / Operator cannot easily enforce a path...

Reachability ≠ Availability

Other challenges Auto-Configuration is important. Protocols should negotiate and come up without any manual configuration... Provisioning can be simplified (lack of standardization) Turning on a network requires another network (out of band) (To hardware vendors) BMC in every switch is a MUST!