Storage class and access behavior

Considering the high latency of tape, glacier storage class capability is deployed in IBM Deep Archive. More specifically, only Glacier Flexible Retrieval storage class is supported on Deep Archive.

When S3 APIs such as S3 PutObject, S3 CopyObject, and S3 CreateMultipartUpload are used to upload an object, the Glacier Flexible Retrieval storage class must be specified. If the storage class is not specified, the upload S3 APIs fail. As defined in Amazon S3 API reference version 2006-03-01, the value to specify for the x-amz-storage-class header is GLACIER.

The storage class of an object is always Glacier Flexible Retrieval. Any S3 APIs that change the storage class are not supported. Also, the storage class does not change regardless of whether the object data is in cache or tape in IBM Deep Archive.

When an object is uploaded by the S3 API, the uploaded data is first stored in the cache in IBM Deep Archive. The response of the S3 API is returned, and within a maximum of 10 minutes, an internal process is triggered to migrate the data from the cache to tape.

To download an object, the object must first be restored by the S3 RestoreObject API. Using the S3 GetObject API without requesting the restore operation by the S3 RestoreObject API fails. The S3 RestoreObject API is an asynchronous API, and the response is immediately returned. Within a maximum of 10 minutes, an internal process is triggered to recall the data from tape. The status of the restore operation can be queried by the S3 HeadObject API.

Table 1. Header values
x-amz-restore header value	Description
ongoing-request=”false”	A restore request has not been issued to the object.
ongoing-request=”true”	A restore request is accepted.
ongoing-request=”false”, expiry-date="Wed, 5 June 2024 00:00:00 GMT"	The restore operation is complete. The object data is recalled to a temporary cache and is ready for download.
ongoing-request=”false”	The duration is expired, and the temporary cache is deleted.

When the restore is complete, the object data is copied to a temporary cache for the duration specified in the S3 RestoreObject API. The maximum days that are supported for the duration are 30 days. After the completion of the restore, the data can be downloaded by the S3 GetObject API. After the duration is expired, IBM Deep Archive removes the temporary cache data.