1) Tape: Tape, should and in many cases is, a prominent player in the end to end architecture of a cloud storage provider. As much as we love disk, consider that even in the cloud a copy of all data should be stored on tape. If it isn’t, it is at risk for being lost. Permanently. I did not make this up. Just review news stories about cloud outages with lost data. Replication, snapshots, CDP, RAID, and m-of-n protection are great innovations in disk-based data protection. However, they are not enough. Very large data sets push the error rate thresholds of modern storage systems from statistically negligible to a very plausible reality. The short version: not having an isolated, offline copy, implies an inherent risk, and tape is still the best media suited for offline storage.

This copy of data cannot be should not be able to be accessed, changed, or deleted without some form of human intervention or negligence.  With libraries such as Spectra libraries, it’s easy to encrypt the data and store the tape in the library. An encrypted tape stored in an environmentally stable, secure location is the best method for keeping an offline copy. And, as stated many times before, it is still prudent to maintain a copy of your data, regardless of its use model, within your own storage infrastructure.

About tape: Tape, not disk, is designed to be well suited to offline and off-site storage. Yes, if you leave it on a heater, in the sun, in your trunk, or next to your electromagnetic generator it probably won’t restore, but if you did that to your disk, the data wouldn’t restore either. If you use proper data management techniques tape is much more very reliable than disk.

2. Strong Service Level Agreements (SLA): Make sure your cloud agreement includes SLA’s that align with your usage needs. With the cloud, you get what you pay for. That is both the advantage and risk of using cloud-based storage. If you are using the cloud as an availability or distribution system, then standard SLA’s are most likely fine. However, if it is your sole copy, or only backup of your data, make sure you are investing in a storage service designed to protect that data in the event of an issue at the hosting site. You are only protected as much as your SLA agrees to. If it doesn’t commit to getting your data back in the same condition it was sent (many basic SLA’s don’t) then it isn’t well suited for a backup or worse yet primary target of your company’s assets. Expect your data to be available and healthy, but defend yourself against unexpected outages or data loss by knowing what your SLA agreement is. Also, make sure you know your cloud service provider’s data protection strategy. They may not be willing to share every specific vendor used, but methodology can be disclosed without disclosing specific vendors, which in turn will give you a much more accurate picture of how well your data is protected.

3. An Exit Strategy: While the idea is to store data in the cloud, make sure that there is a realistic way to retrieve or migrate your data to another cloud provider or back to your internal systems. This protects your data in the event that either your company discontinues usage or the hosting company discontinues the service.

Further, keep an eye on the amount of data you are storing in the cloud. It is very likely that the amount of data you are storing is very likely to grow over time, and could outgrow the realistic cost/time associated with sending that data across a WAN. Again, tape is an excellent method of handling seeding and exit strategies. Particularly with open formats of tape, such as LTFS or TAR it’s straightforward to transfer data between two heterogeneous environments. In the event that you have hundreds of terabytes or even petabytes, shipping media is often faster and considerably less expensive than paying for the bandwidth required to download that much data. Additionally, in the event that a hosting company goes out of business, open formatted tapes can be distributed even if the entire hosting system is no longer online. It’s just smart to be able to get your data no matter what happens to the host.