About
We have set-up an OpenStack-based cloud infrastructure in the framework of a publicly funded project, PRISMA, aimed at the implementation of a fully integrated PaaS+IaaS platform to provide services in the field of smart-government (e-health, e-government, etc.). The IaaS testbed currently consists of 18 compute nodes providing in total almost 600 cores, 3550 GB of RAM, 400 TB of storage (disks). Connectivity is ensured through 2 NICs, 1Gbit/s and 10Gbit/s. Both the backend (MySQL database and RabbitMq message broker) and the core services (nova, keystone, glance, neutron, etc.) have been configured in high-availability using HA clustering techniques. The full capacity available by 2015 will provide 2000 cores and 8 TB of RAM. In this work we present the storage solutions that we are currently using as backend for our production cloud services. Storage is one of the key components of the cloud stack and can be used both to host the running VMs ("ephemeral" storage), and to host persistent data such as the block devices used by the VMs or users' archived unstructured data, backups, virtual images, etc.. The storage-as-service is implemented in Openstack by the Block Storage project, Cinder, and the Object Storage project, Swift. Selecting the right software to manage the underlying backend storage for these services is very important and decisions can depend on many factors, not only merely technical, but also economic: in most cases they result from a trade-off between performance and costs.
Many operators use separate compute and storage hosts. We decided not to follow this mainstream trend aiming at the best cost-performance scenario: for us it makes sense to run compute and storage on the same machines since we want to be able to dedicate as many of our hosts as possible to running instances. Therefore, each compute node is configured with a significant amount of disk space and a distributed file system (GlusterFS and/or Ceph) ties the disks from each compute…
The talk
Credits and licence
- Institution
- CERN
- Department
- HEP Computing
- Taught by
- Spinoso VINCENZO, Marica ANTONACCI, Giacinto Dr. DONVITO
- Year
- 2014
- Language
- English
- Subjects
- Computer science
- Source
- CERN
- Licence
- CERN, as CERN states it. CERN's terms
- Attribution
- © CERN. CERN provides its audiovisual media free of charge for educational and informational use, credited to CERN; this lecture is shown in CERN's own player.
- Who can watch
- CERN gives no rating. Invisible Network shows it to Tweens profiles and older, judged from what the source says about it. How courses are placed