OpenStack Summit November 2013

Wee from Intel IT Engineer Computing, which is a group to provide Intel private cloud solution. Currently, we have a 36-node OpenStack Cluster (based on Grizzly) providing over 1000 VMs to our internal customers. Our OpenStack journey started from Essex. During this one year, we ran into a lot of problems and resolved them. It time for us to share with the community about our experience.

(1) How to manage version update? I believe it very painful for each half year update. We use Continuous Integration to resolve this problem. We track the source code from github every day. Once the code is changed, our CI server will check out the code and deploy it to our test environment and run all the tests. We add our customized tests into tempest. By using CI locally, we can integrate the lasted code or tagged code daily and resolve the update issue in every dayswork. At the same time, we can track the progress of the community.

(2) Tempest is not perfect. Wee using tempest in our CI environment. However, tempest will cause many issues to your environment after you run it. We have added a new module called tempest launcher to help us clean up the environment after running tempest.

(3) More powerful and friendly portal. We develop a customized portal called Taurus which is an operation one-stop-shop center. Taurus not only covers all the functions Horizon have, but also integrate with nagios & ganglia. You can also use Taurus to trigger Tempest and view its result friendly.

(4) Improve the speed of snapshot. Original snapshot is slow because of rebase and uploading each time. We did some improvement in this feature to reduce the snapshot time, which save our customers lots of time.

(5) One click deployment. Based on dodai, we burn a live customized image into USB. And using this USB, it very easy to deploy an OpenStack Cluster.

(6) BI in cloud data analysis. We launch Hadoop on our OpenStack Cluster to analysis this cluster log data nightly and archive the data into Ceph.