I am designing copy long-term and disaster recovery copies in Amazon S3 and I have doubts about how to design the secondary copies.Primarily copies, for the short-term, are held on-site in a deduplicated disk library for 30-45 days. Restores are very infrequents.
My first approach is to create a secondary copy with deduplication, and send the data to a cloud library configured on an S3 bucket.The cloud library is configured using combined storage class (S3-IA/Deep Glacier). This copy will be a Weekly select copy and then apply extended retentions. Basic retention to 180 days to prevent EarlyDelete on Deep Glacier, monthly retentions to 385 and Yearly to 1825. My first concern is the size of the DDB using extended retentions (CV recommend seal the DDB each 6-12 months so maybe this is not a big problem). Because the use of extended retentions, copies will the retained until the longest retention and producing mixed contents;The advantage I think is that the first baseline will be almost the same for each retention copy.
My second approach is to create a secondary copy with deduplicacion for each retention, so three DBs: one for Weekly, Monthly and Yearly and create three different S3 buckets for each retention.
The three cloud libraries configured using combined storage class (S3-IA/Deep Glacier).
Copies will be selective, each retention per copy. The advantage I think is that the DDB will be sized to the rentetion copy, optimal performance.The disadvantage is the first baseline will be uploaded three times.
The last approach is similar to the last one, but creating only one S3 bucket.
Here I have doubts if when creating three DDBs, and only one cloud library, you will upload the baseline three times, or will upload it only once.
Can someone recommend me the best option given this scenario?