The Eight Fallacies of Distributed Computing

by Peter Deutsch (sent to me by roy)

"Essentially everyone, when they first build a distributed application, makes the following eight assumptions. All prove to be false in the long run and all cause big trouble and painful learning experiences.

1. The network is reliable
2. Latency is zero
3. Bandwidth is infinite
4. The network is secure
5. Topology doesn't change
6. There is one administrator
7. Transport cost is zero
8. The network is homogeneous"

I do make those assumptions when I go distributed but, fortunately, I don't just go distributed for every project I design. for instance, I rarely implement a distributed system to save just a few seconds on operations like UI, data transfer, etc. I only do it for heavy hours-long calculations that can be divided and localized. I employ passive distribution where tasks are not pushed onto the nodes, they are put in a job pool and the nodes just pick from there. a job is only specs for a product (low bandwidth), the logic is preinstalled on the nodes. when a job is not finished within a specific time, it is rendered obsolete and a new job of the same specs will be created with a longer timeout. therefore, items 1, 2, 3, 5, 6 and 7 are taken care of. item 4 is only important if security is an issue. so the only thing I have to ensure is item 8.

Deustch made very good points, and I believe those issues are what naturally drove me to build the kind of distributed systems I do.