The Two Wells: A Parable of Backup Strategy
There are two old wells at the edge of my property, remnants of the farmstead that stood here a century ago. They are both deep, and both find the same clear water from the same deep aquifer. But they are built differently, and this difference has always struck me as the perfect analogy for the two philosophies that govern how we approach one of the most fundamental acts of stewardship in our little technological domains: the backup.
The first well is a classic stone cylinder, capped with a heavy wooden lid and a simple rope-and-bucket mechanism. It is direct and immediate. When you need water, you lower the bucket, it fills, and you haul it up. It’s a single, tangible source. The second well is less obvious. A few yards away, a buried clay pipe runs from the aquifer to a small, constantly-fed stone trough. The water flows ceaselessly, silently filling the basin before spilling over into a drain that returns it to the earth. It is not a single source, but a continuous stream.
Our digital backups often follow the design of the first well. We carve out a point in time—midnight on a Tuesday, for instance—and we lower the bucket. We pull up a complete, monolithic snapshot of our data, label it carefully, and store it away. This is the Full Backup. It is reassuringly solid. You can point to the file, measure its heft, and know that at that precise moment, everything was captured. It is our stone well, a definitive archive. Its strength is its clarity and completeness. When disaster strikes, the recovery path is conceptually simple: haul up that one bucket and pour its contents back into the empty cistern of a new system.
The trough, however, represents the philosophy of the continuous stream: incremental backup, or the modern preference for write-ahead logs and streaming replication. Here, the goal is not to capture a single state, but to record every change as it happens. The data flows from the live system into a journal, a steady trickle of events. Restoring is not a matter of fetching one large bucket, but of replaying the stream from a known good starting point up to the moment before the failure. Its strength is its granularity. If the failure occurs at 3:15 PM, you don’t lose everything since last night's backup; you lose only the seconds or minutes since the last event was safely recorded in the stream.
Choosing between the bucket and the stream is the essential, quiet decision. The bucket demands significant, periodic effort and storage, but offers a quick, monolithic restore. The stream operates with a gentle, constant hum, consuming resources evenly, but can make restoration a more complex journey of reassembly. The bucket is a stark before and after. The stream is a narrative. One is not inherently superior to the other; they are simply different tools for different tolerances of loss and effort. The wise homesteader, I’ve found, doesn't choose just one. They maintain the stone well for its deep, certifiable reserves, but they also appreciate the trough for its constant, life-giving flow, knowing that true resilience often lies not in a single perfect solution, but in the thoughtful layering of contrasting designs.
Notes & further reading
A few pages I came back to while writing this: