Getting Started

We're going to go over some simple terminology and concepts commonly used in Atmosphere.

Terminology - Atmosphere

User

A User is an individual that has logged into Atmosphere.

Users have:

• Username (unique)
• Name
• Email

Group

Groups are used similar to the linux “user/group” model.

• A Group of one will exist for each User.
• A group can contain one or more Users
• Users can be in more than one Group at once
• Groups (not Users) are the “owners” of cloud resources

Provider

Provider (also known as cloud, cloud provider) is a term used by atmosphere to identify a specific installation of cloud software.

• Providers have a specific set of credentials common to all users (That uniquely identifies them)
  • Ex: auth_url, version, …
• ProviderType
  • Ex: OpenStack, AWS, …
• Providers are owned by a Group (usually a staff/admin)

Identity

Identities are a term used by atmosphere to identify a specific user on a Provider.

• Identities have a specific set of credentials for the user
  • Example: key(username), secret(password)
• Identities are owned by a Group

Projects

Projects (in Atmosphere) represent a collection of Cloud Resources (not sizes) that allow users to logically group related instances/volumes.

• A project is required for new instance/volume created in Atmosphere.
• Projects have a name
• Projects are owned by a Group

Application

Applications (displayed to the user in troposphere as Image) are a way to abstract a “history” of related Machines (via ApplicationVersion).

Ex:

ApplicationVersion

ApplicationVersion (also known as Version) is a grouping of one or more Machines.

(In Atmosphere, it is common to replicate a Machine (ex: Ubuntu 16.04) across several Providers (ex: iPlant - Austin, iPlant - Tucson, …).)

Because the image is (mostly) identical across providers, the grouping serves as a way to simplify the cloud-agnostic nature of Atmosphere.

• ApplicationVersions have One or more ProviderMachine
• ApplicationVersions have a name (1.0, 2.0-alpha, beta-do-not-use, …)
• ApplicationVersions have a change log (that helps users understand what changed between versions)

Terminology - Cloud

Cloud resource

Cloud Resources are a term used by atmosphere to identify a Volume, Instance, Size, or Machine.

Size

Sizes (also known as flavors in Openstack) are a preset configuration of CPUs, GB of memory, and Disk space that will be allocated to a Machine when creating an Instance.

Sizes have:

• CPU
• RAM (in GB)
• Disk (for the image, in GB, 0 means "create disk image that is the exact size of the Machine")
• Ephemeral (as an additional mount, in GB)

Machine

Machines (also known as Image, or ProviderMachine) are like a blueprint that will tell the Cloud how to create a new Instance.

• Machines, minimally, have an operating system
• They may also be installed and configured with additional software applications as desired.
• They are a fixed size (usually 8GB-60GB)
• Administrators and other users can create new machines from Instances via Machine Requests and/or downloading new image files.

Volume

Volumes are how users can allocate additional data storage (by policy, up to 1TB) to be used with their instances.

• Volumes have a pre-set size (1-1000gb)
• Volumes start unformatted
  • Atmosphere handles mkfs on first attach + mounting.
• Can be created based on an Image and “Booted” as an instance. (Bootable volumes are not yet a part of atmosphere)

Instance

Instances (also known as nodes, or servers) is the virtual machine that will be created by end users.

• Instances have a specific size
• Instances are based on an Image or Volume.
• Atmosphere handles the networking and pre-configuration of instances prior to handing control over to the User.

Concepts in Atmosphere and Openstack

Concept: Quota and Allocation

Quotas are an admin-set limit on the number of resources a specific cloud user can create at any one time. - (Ex: Quota of 2 CPU, 8 GB ram) Allocations are a way for Atmosphere to track the amount of time a specific cloud user has used their resource. - (Ex: Allocation for 168 hours of compute in one month)

Conecpt: Machine Requests (Imaging)

When a user launches an instance and configures it to their liking, they can create a MachineRequest that will alert staff users to their desire to create a new Machine. Staff members can then accept/reject the request which will start an automated, asynchronous "imaging" process.

During the imaging process: - The 'machine' is created: - Create and Download a snapshot of the Instance - Mount and clean the image file to remove “User-specific” data - Upload new image to cloud - The 'machine' is processed: - Appropriate metadata is attached to the new image - Atmosphere properly catalogs the image into the right Application, Version, and ProviderMachine. - The 'machine' is validated: - Atmosphere will launch the instance and wait for it to go to "green-light active". - Atmosphere will then destroy the launched instance - The 'machine' is completed: - The 'machine owner' is e-mailed - The machine is ready for launch by others.

Concept: Projects

Project is an overloaded term, at any one time it could be referring to: - Atmosphere Project : A collection of Cloud Resources. - OpenStack “project” aka Tenant: A grouping of users inside OpenStack - TACC/Xsede project: A collection of TACC/Xsede users.