# Under My Thumb: Insight Behind the Rules

DevFeed: [Under My Thumb: Insight Behind the Rules](<https://devfeed.tech/articles/under-my-thumb-insight-behind-the-rules-12530.md>)

Original publisher: [Read original article](<http://brooker.co.za/blog/2022/12/15/thumb.html>)

Author: Marc Brooker

Published: 2022-12-15T00:00:00Z

Content type: article

Language: en

Sources: [Marc Brooker's Blog](<https://devfeed.tech/sources/marc-brooker-s-blog.md>), [Marc Brooker's Blog](<https://devfeed.tech/sources/marc-brooker-s-blog-2.md>)

Topics: [systems](<https://devfeed.tech/topics/systems.md>), [data](<https://devfeed.tech/topics/data.md>), [Hardware](<https://devfeed.tech/topics/hardware.md>), [Optimization](<https://devfeed.tech/topics/optimization.md>)

Tags: [cost](<https://devfeed.tech/tags/cost.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [data](<https://devfeed.tech/tags/data.md>), [developers](<https://devfeed.tech/tags/developers.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [memory](<https://devfeed.tech/tags/memory.md>), [optimization](<https://devfeed.tech/tags/optimization.md>), [systems](<https://devfeed.tech/tags/systems.md>)

## AI overview

The article examines how rules of thumb in radar and computer systems can be misleading when removed from their context, while still providing useful starting points for quantitative reasoning. Using Jim Gray's five-minute rule as an example, it explains how comparing memory, storage, CPU, channel, and access costs can estimate when data should remain in main memory.

## Source excerpt

Under My Thumb: Insight Behind the Rules My left thumb is exactly 25.4mm wide. Starting off in a new field, you hear a lot of rules of thumb. Rules for estimating things, thinking about things, and (ideally) simplifying tough decisions. When I started in Radar, I heard: the transmitter makes up three quarters of the cost of a radar system and when I started building computer systems, I heard a lot of things like: hardware is free, developers are expensive and, the ubiquitous: premature optimization is the root of all evil. None of these things are true. Some are less true than others2. Mostly, they're so context dependent that stripping them of their context renders them meaningless. On the other hand, heuristics like this can be exceptionally valuable, saving us time reasoning things through from first principles, and allowing rapid exploration of a design space. Can we make these truisms more true, and more useful, by turning to them into frameworks for quantitative thinking? The 5 Minute Rule Jim Gray's famous 5 minute rule, from 1987: Data referenced every five minutes should be memory resident. Today, thirty five years later, Gray's five minute rule is just as misleading as the ones above1. What we're left with isn't a rule, but a powerful and durable insight. Gray and Putzolu's observation was that we can calculate the cost of something (storing a page of data in memory) and the cost of replacing that thing (reading the data from storage), and quantitatively estimate how long we should keep the thing. They did it like this: The derivation of the five minute rule goes as follows: A disc, and half a controller comfortably deliver 15 random accesses per second and are priced at about 15K$ So the price per disc access per second is about 1K$/a/s. The extra CPU and channel cost for supporting a disc is 1K$/a/s. So one disc access per second costs about 2K$/a/s. A megabyte of main memory costs about 5K$, so a kilobyte costs 5$. If making a 1Kb data record main-memor