iNNOVO Cloud Can You Read this? Red Blue Magenta White Yellow Green -All colors visible? Can You Read this? -All circles round? Can You Read this? -All edges Can You Read this? visible? Can You Read this? Beamer Testpicture 16:9 1 1
The Internet™ 192.168.178.1/24 DHCP 192.168.178.42/24 GW: 192.168.178.1 …
The Datacenter™ 10.23.0.46/24 The Internet™ 10.23.0.45/24 10.23.0.1/24 10.23.0.44/24 10.23.0.43/24 10.23.0.42/24 GW: 10.23.0.1 …
The Datacenter™ Datacenter FeaturesRisks • Portchannels/-aggregation 10.23.0.46/24 The Internet™ • VLANs 10.23.0.45/24 • Spanning-Tree 10.23.0.1/24 10.23.0.44/24 • QoS 10.23.0.43/24 10.23.0.42/24 GW: 10.23.0.1 …
The Datacenter™ 10.23.0.46/24 The Internet™ 10.23.0.45/24 10.23.0.1/24 10.23.0.44/24 10.23.0.43/24 10.23.0.42/24 GW: 10.23.0.1 …
Operating Openstack on an IP-Fabric Jan.Walzer@innovo-cloud.de twitter://@janwalzer 2 0 . N O V E M B E R 2 0 1 8
... a little Agenda What is an IP-Fabric Why Deployment Operations Cumulus 7
To Recap 10.23.0.1/24 10.23.0...../24 10.23.0.42/24 8
10.90.0.1/30 10.90.1.1/30 10.90.0.2/30 10.90.1.2/30 10.90.2.1/32 10
Terminology We call these the transfer-networks 10.90.0.1/30 10.90.1.1/30 We call this the 10.90.0.2/30 10.90.1.2/30 canonical IP 10.90.2.1/32 11
Transfer-Networks 10.90.0.1/30 10.90.1.1/30 The IP-transfer networks only allow 2 IPs to be used There is NO default Gateway 10.90.0.2/30 10.90.1.2/30 10.90.2.1/32 12
BGP – Border Gateway Protocol 10.90.0.1/30 10.90.1.1/30 Routing information All devices have a is exchanged via BGP-Daemon BGP running 10.90.0.2/30 10.90.1.2/30 10.90.2.1/32 13
BGP ... ... 10.90.2.1 via 10.90.0.2 10.90.2.1 via 10.90.1.2 ... ... 10.90.0.1/30 10.90.1.1/30 10.90.0.2/30 10.90.1.2/30 ... 0.0.0.0 via 10.90.0.1 0.0.0.0 via 10.90.1.1 ... 10.90.2.1/32 14
BGP 10.90.0.5/30 ... ... 10.90.1.5/30 10.90.2.1 via 10.90.0.2 10.90.2.1 via 10.90.1.2 ... ... 10.90.0.2/30 10.90.1.2/30 ... 0.0.0.0 via 10.90.0.1 0.0.0.0 via 10.90.1.1 ... 15 10.90.2.1/32
BGP 10.90.0.5/30 ... ... 10.90.1.5/30 10.90.2.1 via 10.90.0.2 10.90.2.1 via 10.90.1.2 10.90.2.2 via 10.90.0.6 10.90.2.2 via 10.90.1.6 ... ... 10.90.1.6/30 10.90.0.2/30 10.90.1.2/30 ... ... 10.90.0.6/30 0.0.0.0 via 10.90.0.5 0.0.0.0 via 10.90.0.1 0.0.0.0 via 10.90.1.5 0.0.0.0 via 10.90.1.1 10.90.2.1 via 10.90.0.5 10.90.2.2 via 10.90.0.1 10.90.2.1 via 10.90.1.5 10.90.2.2 via 10.90.1.1 ... 16 ... 10.90.2.1/32 10.90.2.2/32
17
18
19
Consistent performance, subscription and latency between all racks Consistent performance and latency with scale Architecture for any-to-any Datacenter trafficflows 20
21
Why Operational Flexibility Redundancy& Resilience Scalability 23
Why Every IP in the Cluster can be Operational Flexibility moved to any Server instantly Redundancy& Resilience Scalability 24
Why Every IP in the Cluster can be Operational Flexibility moved to any Avoid Layer2 issues Server instantly Redundancy& L2 Fault Isolation Resilience Multipathing Anycast Scalability 25
Why Every IP in the Cluster can be Operational Flexibility moved to any Avoid Layer2 issues Server instantly Redundancy& L2 Fault Isolation Resilience Multipathing The Design stays the Anycast same! Scalability • Just add more links • Just add more switches • Just add more racks • Just add more datacenters 26
Deployment Tools IP Management: Source of truth Infrastructure as Code Serverconfig Switchconfig 27
Deployment Tools knows the truth ... IP Management: Source of truth Infrastructure as Code Serverconfig Switchconfig 28
Deployment Tools knows the truth ... IP Management: Source of truth Ansible playbooks Infrastructure as Code Toolchain Serverconfig Switchconfig 29
Deployment Tools knows the truth ... IP Management: Source of truth Ansible playbooks Infrastructure as Code Toolchain There is no difference • Switches running Cumulus Serverconfig Linux • No operational difference • Applying the same roles to all devices Switchconfig 30
Operations – Everything is a metric Metric based ... Monitoring Alerting Logging 31
Operations – Everything is a metric Metric based ... Monitoring Alerting Logging 32
Operations – Everything is a metric Metric based ... Monitoring Alerting Logging 33
Operations – Everything is a metric Metric based ... Monitoring Prometheus Alertmanager Alerting Logging 34
„ What makes Cumulus Linux so special ?“
„ What makes • Just a normal Linux Server based on Debian Linux Cumulus Linux • with lots of network interfaces so special ?“ • deployment and configuration • Troubleshooting: regular Linux environment • Monitoring/Logging: Prometheus/Elastic
„ What makes Cumulus Linux so special ?“ The special thing is: Our routers are nothing special anymore!
Questions Jan.Walzer@innovo-cloud.de twitter://@janwalzer
iNNOVO Cloud We are hiring: jobs@innovo-cloud.de 2 0 . N O V E M B E R 2 0 1 8
Backup Slides
DC2 DC1 100G 100G 25G
BGP AS-Numbering DC2 DC1 AS65001 AS65003 AS65002 AS65005 AS65004
Crayon- Icons are cool … http://www.visguy.com/2008/08/11/crayon-network-shapes/
Recommend
More recommend