Challenges of scaling a upstream Openstack Cloud

on YouTube ↗18m

" summary

Société Générale runs an upstream OpenStack private cloud with 18k VMs across 4 regions and ~450 compute hosts (≈350 in Paris), sees ~1,500 instance create/destroy ops per day, and grows control nodes from 3 to 12 in Paris to handle control-plane load. Major problems: massive VLAN/provider-network segmentation causing exploding security-group rules (~160k rules), Open vSwitch agent being single-threaded and failing with ~60–70 rules per port, and segments/subnet limits. Remedies: merged restart-rate patch, segment plugin patch under review, internal security-group and OVS patches planned for upstream.

AI-written summary · about

" chapters (7)

1. 0:00Introduction
2. 0:42Multicloud strategy
3. 1:40Global deployment
4. 3:05Storage
5. 3:43People
6. 9:23Security
7. 12:33QA

" related talks

• 18mAchieve a single Openstack deployment in multiple datacenterVictor Coutellier · OpenInfra Summit 2020
• 30mOpenStack Works, so now what?Monty Taylor · PyCon Australia 2015
• 27mLessons Learned from a Large Scale OpenStack Deployment with TripleoSai Sindhur Malleni, Pradipta Sahoo · OpenInfra Summit 2020
• 19mHow open source contribution has increased work efficiency of a French bank ?Victor Coutellier · OpenInfra Summit 2020
• 34mHeadline Panel: Global Enterprise IT Companies Supporting OpenStackRuchi Bhargava, Mats Karlsson, Mark McLoughlin, Haiying Wang, Mark Muehl · OpenStack Summit Paris 2014
• 22mMoving a Large Workload from a Public Cloud to an OpenStack Private Cloud...Nicolas Brousse · SREcon16 Europe
• 17mBuilding the Rackspace Open Cloud with XenServer and OpenStackPaul Voccio · Xen Project User Summit 2013
• 19m10 years of OpenStack at CERN (From 0 to 300k cores)Belmiro Moreira · OpenInfra Summit 2020