Service Level Description
General Service Description
The CERN Private Cloud provides an OpenStack-based Infrastructure-as-a-Service solution integrated with CERN's computing facilities. Using self service portals or cloud interfaces, users and services can request virtual machines for test, development and production purposes, as well as persistent storage (volumes, object stores, file shares) that can be attached to these VMs. The machines can be of different capacities and run a variety of Windows or Linux operating systems.
Service Interventions and Scheduled Downtimes
Planned interventions that render virtual machines unreachable (e.g. they need to be stopped) will be announced by mail to the VM owner at least one week in advance. For VMs in Personal Projects the time of the intervention is not negotiable. For VMs in Shared Projects, the time of the intervention can be re-scheduled under exceptional circumstances only (please get in touch with the Cloud team via the CERN Service Desk in that case).
General service interventions will be announced at least one week in advance on the IT Service Status Board.
Default Quotas for Cloud Resources
The amount of resources available to a user (in a personal project) or a service/activity (in a shared project) are controlled by a quota system. The current default quotas are
| Type | Personal Project | Shared Project |
|---|---|---|
| Instances | 5 | 25 |
| Cores | 10 | 25 |
| RAM [GB] | 20 | 50 |
| Volumes | 10 | 5 |
| Volume Space [GB] | 100 | 500 |
Following an approval process, the quotas of shared projects can be changed. In order to have the quota of a shared project changed, please file a Quota Change Request.
Estimated Virtual Machine Availability and Data Durability
It is expected that the availability of virtual machines provided by the service will be comparable to the availability of the underlying physical hardware. Taking hardware failure rates into account, it is estimated that the service will lose about 1 VM per 5000 per year.
For the durability of uploaded images and data on persistent storage (volumes), please refer to the Service Level Description of Ceph as the main backend store.
Performance
Since the service comprises a variety of different hardware (in terms of processors, network interconnects, or disk technologies) and some of these resources in general are shared or may be overcomitted, there are no guaranteed minimum performance levels given here. If your application requires certain performance levels and the performance you experience does not meet these, do not hesitate to get in touch with the Cloud team via the CERN Service Desk.
Backups
The cloud service does not offer automatic backups of the virtual machines or their data. Please refer to the Backups and Snapshots for the options to configure backups of your virtual machines and their data.
Support
User support during CERN working hours is provided by the CERN Service Desk. Support outside working hours is best effort only.
Note: Members of WLCG Virtual Organizations can raise GGUS ALARM tickets for urgent production problems.
User Responsibilities
Requestors of VMs need to ensure the following actions are performed:
- Comply with the CERN Computer Security Rules
- Arrange prompt update of operating system patches. By default, the standard images provided by IT will automatically update with yum and CMF, respectively.
- Provide timely responses to mails from the IT service teams regarding service issues and requests for information.
- If user defined images are uploaded, the user is responsible to keep a copy of the image in case of accidental deletion or corruption.
- Delete their VMs when no longer needed.
Annual closure
During the CERN annual closure at the end of each year around Christmas, there will be only best effort support for the cloud service. In particular, no hardware interventions will take place.