I have long considered the cloud better suited to data than apps. I can’t explain my aversion to web apps, the cloud just seems better suited to serving data than CPU time.
We access the data on our computers via a hierarchical caching structure that grows successively larger and slower as data travels farther from the CPU. We only access a small set of this data frequently, such as the contents of the file we have open, but can afford to store the majority in low-cost, slow memory. Until recently, the lowest and slowest rung of the hierarchy was our hard drive — a mechanical device optimised for capacity above all else. But there is a fundamental limit to the lifespan of hard drives — all fail eventually. As precious mementos take up permanent residence on these fickle devices, we become vulnerable to data loss. Cloud backups are an excellent solution to this problem.
To operate effectively as the next level of the data heirarchy, however, cloud storage should provide:
- Automated transmission of incremental changes
- Some means of versioning and rolling back changes
- Reliable, secure, and remote access
Each of these features requires both remote storage and local client software. I have tried several such services over the years, including:
- Dropbox,
- Crash Plan, and
- Arq
While writing my thesis, I relied heavily on Dropbox to backup, version, and synchronise my files across devices. It was elegant and simple, but costly at $15/month. Moreover, it seems optimised for data synchronisation rather than cheap, reliable backups. Later I learned that Dropbox was insecure by design. Unsettled by my reliance on Dropbox, I started using Crash Plan, who provide ‘Trust No One’ (TNO) security for $3/month, where all data is encrypted using a local key before being uploaded to the cloud. The only downside of Crash Plan is that it runs in Java, and I hate Java with the fire of a thousand suns. I hate it more than web apps. In fact, I hate Java more than I hate pineapple on pizza, which is pretty much the limit of a mankind’s capacity to hate.
And so I found Arq, which is TNO, uses an open-source encryption algorithm, and uploads data to Amazon Glacier for extremely cheap storage ($0.30/month for my 30GB of data). It works in the background, consolidating, compressing and transmitting incremental versioned backups, and it runs natively on my Mac.
Admittedly, retrieving data from Amazon Glacier is slow and costly, but since I won’t retrieve the data unless all local copies are inaccessible, I consider this an acceptable trade-off. The only downside is that Arq is not platform agnostic, and I am nervous about locking myself into a single platform. I am still, on balance, happy as a Mac user, but Apple’s software appears to be getting both buggy and dumbed down to the point of being unusable. If I find a cloud storage service that addresses both gripes, I may switch, but for now, Arq is my cloud backup service of choice, and I highly recommend it.