Austin
April 25-29, 2016

Event Details

Please note: All times listed below are in Central Time Zone


Performance Analysis in Large-scale Deployment - A Single Thousand-nodes Cluster

How to desigin the architechture of deploying OpenStack in a single thoudsand-nodes cluster? How much pressure would basic component such as database, messaging queue and loadbalancer encounter?Where is the bottleneck? Here, we have built a single 1024-nodes OpenStack cluster which contains about 30 controller nodes,  650 compute nodes, 250 storage nodes and 20 network nodes in China Mobile Public Cloud.  We did a lotof performance tests as well as recording  the data of each component in detail.  Based on these data, we did careful analysis and lots of improvements.  We share them  to  all OpenStackusers and hope they would help you optimize the architecture and improve the performance  in large-sale deployment!


What can I expect to learn?

The deployment architecture of  thousands of nodes in single OpenStack cluster.The performance of  1024-nodes in single OpenStack cluster.The monitoring data of MySQL, RabbitMQ and Haproxy in performance tests.Data Analysis and our improvements.

Tuesday, April 26, 2:50pm-3:30pm (7:50pm - 8:30pm UTC)
Difficulty Level: Beginner
senior engineer
Luo Gangyi, Data of Birth: 1989.03,Education: Master in Computer Science, Nanjing University. He is currently working in China Mobile as a software engineer and mainly doing IAAS software development. Also he is a contributor to Openstack Community. FULL PROFILE
Senior Engineer
I graduated from North China Electric Power University. Undergraduate major is software engineering and graduate professional is computer software and theory. During school days I was committed to virtual desktop infrastructure. At present I worked in China Mobile and engaged in openstack nova, glance, cinder and ironic development and maintenance. FULL PROFILE
Comments
0 Reviews
0