End-to-end fabric visibility Bringing web-scale networking September 21 th, 2017 Attilla de Groot SE, HCIE #3494 1
Cumulus Linux architecture Uniform operating model write any tools, use any apps User Space Routing Suite Network Orchestration Bridging VXLAN Automation Monitoring Third Party/Customer Applications switchd Switch HAL Linux Kernel VxLAN Routing Tables ARP Table Bridge Table Ethernet Interfaces ASIC Driver CPU, RAM, Flash, etc. Switch Silicon Front Panel Ports 2
& Open source Creating new technology Open networking concept ONIE APD Linux networking development VRF Quagga / Free Range Routing Ifupdown2 Iproute2 3
IP-Fabric IP-Fabric with Overlay Clos fabric ECMP over spines MLAG to servers Routing protocol to TOR VxLAN encapsulation EVPN Controlplane MLAG ebgp / OSPF ECMP MLAG 4
Full L3 Fabric with Hostpack Routing to the Host IP Fabric to the host No more L2 domains Free from proprietary protocols FRR support ebgp / OSPF ECMP ebgp / OSPF ECMP ebgp / OSPF ECMP FRR 10.1.1.1/32 Lo1 FRR 10.1.1.2/32 Lo1 FRR 10.1.1.3/32 Lo1 FRR 10.1.1.4/32 Lo1 5
Bridge Network Bridge Network Local bridge Subnet per host NAT / PAT for containers Iptables security 6
Macvlan Macvlan Host interface split Redistribute neighbor 7
Containers in the routing domain Routing on the host Free range routing BGP CRoHDaD Local Docker API Route redistribution 8
Network Blind Spots in Containerized Environment What happens with issues of application availability, performance problems or outage? NetOps Knows the network designed for reliability Does not necessarily know the applications Haven t architected network for rapid change VLAN10 VLAN10 VLAN20 VLAN20 NO VISIBLITY VTEP VTEP VTEP VTEP VTEP VTEP VTEP VTEP VTEP VTEP VTEP VTEP VTEP VTEP VTEP VTEP DevOps Knows applications designed for agility Does not necessarily know the network Containers are spun up and torn down rapidly 9
Traditional Network Tools LIMITED VISIBLIITY REACTIVE MANUAL 10
Can't keep up with the speed of automation CONFIG CHANGES PUSH TO PRODUCTION REACT TO TICKETS/ISSUES MANUAL TROUBLESHOOTING BOX-BY-BOX >_! 11
Can't keep up with the speed of automation CONFIG CHANGES PUSH TO PRODUCTION REACT TO TICKETS/ISSUES MANUAL TROUBLESHOOTING BOX-BY-BOX >_! 12
Can't keep up with the speed of automation CONFIG CHANGES PUSH TO PRODUCTION REACT TO TICKETS/ISSUES MANUAL TROUBLESHOOTING BOX-BY-BOX >_! 13
Can't keep up with the speed of automation CONFIG CHANGES PUSH TO PRODUCTION REACT TO TICKETS/ISSUES MANUAL TROUBLESHOOTING BOX-BY-BOX >_! 14
Can't keep up with the speed of automation CONFIG CHANGES PUSH TO PRODUCTION REACT TO TICKETS/ISSUES MANUAL TROUBLESHOOTING BOX-BY-BOX >_! Existing tools do not offer closed-loop feedback, Adding more of these tools doesn t solve the problem! 15
NetQ brings web-scale efficiencies to network operations Design Provision Automate Validate Troubleshoot Analyze Day 0 to day 1 Day 2 and on Cumulus Linux Open Network Operating System Open OS foundation of network flexibility & choice of apps Disaggregation of SW & HW supply chain freedom Speed of provisioning workloads with automation at scale Cumulus NetQ Telemetry-Based Fabric validation application Validate changes before & during production rollout Precise fault location alerts for rapid problem resolution Time-machine diagnosis & troubleshooting 16
NetQ Deployment Streaming Telemetry 17
NetQ: How it Works NetQ BGP mlag VxLAN Linux Kernel NetQ Agent Subscribes to Linux Kernel Events Interface State MTU Routes, MACs NetQ Agent Polls Routing Information BGP and OSPF Neighbor States Error Conditions 18
NetQ: Fabric Change Log NetQ New Route Added BGP mlag VxLAN OSPF Neighbor Change MAC Address Removed Linux Kernel See state now or any point in the past 19
NetQ: Analysis Engine Validate Current State BGP OSPF MTU mlag VxLAN Telemetry Server analyzes entire network state 20
NetQ: Intelligent Visibility View remote information IPs MACs OS System Specs Improve Command Outputs Resolve hostnames in any Linux command No need for DNS 21
NetQ: Advanced Notification NetQ Notifier Service Automatically Alert on Check Failures Syslog ChatOps (Slack) ELK Splunk Pagerduty 22
NetQ & Container Demo 23
24
Cumulus in the Cloud CITC One click environment No self hosting Testing automation Experience NetQ 25
Cumulus Whitepapers Container networking introduction: https://cumulusnetworks.com/learn/web-scale-networkingresources/white-papers/containers/ Validated design guide: https://cumulusnetworks.com/learn/web-scale-networking- resources/validated-design-guides/validated-design-guide- Container-Networking-Addressing-Routing-Domain/ 26
Join the Community! Get involved in the conversation: Join the community & share your Cumulus Linux tips and experiences: https://getsatisfaction.cumulusnetworks.com/cumulus Chat with others on our Slack channel: https://cumulusnetworks.slack.com/ Keep on learning: https://community.cumulusnetworks.com Download Cumulus VX to work on proofs of concept, test your automation and continue practicing: https://cumulusnetworks.com/cumulus-vx/ Learn at your own pace with demos using our reference topology: https://github.com/cumulusnetworks/cldemo-vagrant https://cumulusnetworks.com/products/cumulus-in-the-cloud/ 27
Thank you! 28