File System based Bank implementation
Currently we suppport Swift and File System as bank implementations. We should increase more bank plugin types so that users will have more choices to feet their needs in different scenarios.
S3 compatible storage is a valid choice, which is used by many individuals and companies in the public or private clouds. With S3 based implementation, it will store objects and object metadata on S3 compatible Storage.
As explained, deployers who want or will use S3 compatible storage in their cloud.
Objects would be stored under a object name with their ID having / be defined as a separator.
- For example::
Object ID: /checkpoints/2fd14f87-46bd-43a9-8853-9e1a84ebee3d/index.json
The metadata files will be in a JSON format. The name and format of these files are same as the meatadata objects in the Swift bank.
- For example::
/checkpoints/3a4d76e7-f8d8-4f2f-9c1d-107d88d7a815/ <- directory /checkpoints/3a4d76e7-f8d8-4f2f-9c1d-107d88d7a815/metadata <- md file /checkpoints/3a4d76e7-f8d8-4f2f-9c1d-107d88d7a815/status
Do nothing, this is not a mission critical feature.
Amazon S3 REST API Introduction * http://docs.aws.amazon.com/AmazonS3/latest/API/Welcome.html
The python client module that could be used is botocore * https://github.com/boto/botocore
Data model impact
REST API impact
Other end user impact
This API might be fasterslower than Swift depending on use case.
Other deployer impact
Primary assignee: Pengju Jiao <email@example.com>
- Write Bank Plugin
- Add documentation
Unit tests in Karbor.
New docs to explain how to use and configure the alternative Bank implementation.