Cloud tiering

Cloud tiering moves cold data from on-premises storage to cloud storage classes priced for infrequent access. The savings are real, and the risk is read cost, because cloud providers charge more to read data out than to put it in. The decision depends on access history, not on capacity.

Key takeaways

Cloud storage classes for infrequent access cost less to store and more to read.
Read and egress charges decide whether a move saves money.
A cloud storage gateway changes where files live, not how they are described.

What is an example of a cloud tiering rule?

Move files in finished project folders that nobody has opened in a year to a cloud archive tier, keep camera originals and masters on premises, and never move files under legal hold. The rule runs once a month and records every file it moves, so users can still find each file and ask for it back.

How do egress costs affect cloud storage decisions?

Cloud providers generally charge little to put data in and more to take it out. That makes the cost of a tiering decision depend on how often the data will be read afterward, so archive tiers with low storage prices can cost more overall if access patterns were misjudged. Last access history is what predicts that.

What is a cloud storage gateway?

A cloud storage gateway is a device or service that presents cloud object storage to local applications as a file share, usually over Network File System (NFS) or Server Message Block (SMB). Applications keep reading and writing files, while the gateway stores the data in the cloud and keeps recently used data in a local cache. It makes the cloud reachable, and it does not describe what the files are, so a catalog is still needed to decide which files belong there.

How do you decide what data to move to the cloud?

Start from what the metadata says about each file, not from file age alone. Four checks cover most decisions:

Access history: files that no one has read in months are the safest candidates, because they are unlikely to be read back.
Business attributes: attributes such as project status, owner, and department show which data belongs to finished work and which still belongs to active teams.
Recall cost: estimate how much data will come back and how often, since retrieval and egress fees can cancel the savings.
Rules on location: retention, compliance, and data sovereignty requirements can limit which data is allowed to leave the data center.

What data should stay on premises?

Data that is read often, by many people, or by applications that need fast and steady response times usually stays close to the users and compute that depend on it. Active projects, large working sets that are read back again and again, and data that contracts or regulations keep in a specific location are common examples. For these files, frequent recalls and egress fees can make the cloud cost more than the local storage it replaces.