Cloud: Difference between revisions

From Alliance Doc
Jump to navigation Jump to search
(Added Cedar login URL)
No edit summary
Line 51: Line 51:
* Some [[VM Best Practices|best practices]].
* Some [[VM Best Practices|best practices]].
* Describe your VM infrastructure as code using [[Terraform]].
* Describe your VM infrastructure as code using [[Terraform]].
* [[Using cloud gpu|GPU Driver Installtion]]


= Cloud systems = <!--T:8-->
= Cloud systems = <!--T:8-->

Revision as of 20:00, 9 March 2020

Other languages:

The Compute Canada cloud resource or "CC-Cloud" is a pool of hardware supporting virtualization. This can be thought of as Infrastructure as a Service.

A user of the cloud will typically create or "spin up" one or more virtual machines (VMs or "instances"). He or she then logs into the VM with administrative privileges, installs any desired software, and runs the software applications needed. These applications could be as diverse as a CPU-intensive analysis of particle physics data, or a web service directed towards scholars of literature and the humanities. The advantage is that the user has complete control over the collection of installed software (the "software stack"). The disadvantage is that the user must have some degree of experience in installing software and otherwise managing a computer.

Virtual machines can be easily replicated. One can take a "snapshot" of a VM which can then be started again elsewhere. This makes it easy to replicate or scale up a service, and to recover from (for example) a power interruption.

If you can fit your work easily into the HPC batch submission workflow and environment (see What is a scheduler?) it is preferable to work outside the cloud, as there are more resources available for HPC and software is already configured and installed for many common needs. There are also tools like Singularity to run custom software stacks inside containers within Compute Canada HPC clusters. If your need isn't served by Singularity or HPC batch, then the cloud is your solution.

Getting a Cloud project[edit]

  • If you do not have a Compute Canada account, create one with these instructions.
  • If you are a primary investigator (PI) with an active cloud resource allocation (see RAC) you should already have a project. See the below section using the cloud to get started. If not or you are not sure please contact technical support.
  • Otherwise go here to:
    • request access to an existing project
    • and if you are a PI you may also:
      • request a new project with our Rapid Access Service (RAS)
      • or an increase in quota of an existing project.

Requests are typically processed within two business days.

Preparing your request[edit]

  • When requesting access to an existing project you will need to know the project name and type (compute or persistent). See the section on projects for guidance on how to find this information. Requests for access must be confirmed by the PI owning the project.
  • When requesting either a new project or an increase in quota for an existing project some justification, in the form of a few sentences, is required:
    • Why you need cloud resources
    • Why an HPC cluster is not suitable
    • Your plans for efficient usage of your resources
    • Your plans for maintenance and security (refer to this page)
  • A PI may own up to 3 projects, but the sum of all project quotas must be within the RAS allocation limits. A PI may have both compute and persistent cloud RAS allocations.

Using the Cloud[edit]

Cloud systems[edit]

The details of the underlying hardware and Openstack versions are described on the cloud resources page. The System status wiki page contains information about the current cloud status and future planned maintenance and upgrade activities.


Support[edit]

For questions about the Compute Canada cloud service, contact technical support.