Introduction to Cloud Computing for Geosciences

There are several different cloud service models.

Deployment models:

• Private

  • Managed locally without shared resources.

  • Useful when security is the key issues

  • Be aware of the sustainability, maintainability, and stability.

• Public

  • Services provided by third party.

  • Provider hosts multiple clients.

• Hybrid

  • Could the “best of both worlds” but could also be “worst of all worlds”.

  • Leverage the public cloud to handle high source demand.

• Community

  • “Private cloud”++

Storage:

• Three key properties

  • instance, volume, and object.

• Instance Storage

  • “Ephemeral storage,” but speed is an advantage.

• Volume (Block) Storage

  • Most like a traditional hard drive.

• Object Storage

  • Limitless space.

  • Backable is available, but cost is associated.

• Several use cases are discussed:

  • Ex: Batch processing, data locality, data-as-a-service

Cloud Economics

• Cost saving is more favorable with performing larger amount of tasks.

• Different versions of pricing: on demand (pay as you go), reserve (flat), and spot (fluctuates, similar to stock market prices).

Caveats:

• Not all clouds offer the same resources/databases.

  • Moving between clouds could be difficult.

• Associated cost could be a concern.  

• Developer paradigm shifts.

  • Will not be sensible to reinvent tools.

• Scalability can also be challenging.

Questions:

• How does cloud service deal with reliability and redundancy?

  • It is important to plan for both elements; it should not taken for granted or considered to be inherent of the cloud technology.  Making backup, redundant copies is still very important.

Identifying and understanding the types of users and the support resources that would be available, such as project coordination and security, are important when choosing the type of cloud services.

• A live demo for Spatiotemporal Cloud Computing Facility was provided regarding on a team could work together to create, manage, and support cloud activities via a dashboard.

• The ESIP cloud computing cluster will announce this fall when the Spatiotemporal Cloud Computing Facility is available as a resource.