Tuesday, November 15, 2016

Methodology done RIGHT

"Pick your battles"
Some battles are a lost cause. Decimate is beyond saving. The historically specific meaning about the Roman practice will probably stick around in specialized discussion, but the proper general meaning of "a 10% loss" isn't coming back.

The same is probably true of methodology and problematic. I only complain about these because I see them used incorrectly all the time. I'm being a hypocrite here because I am a bad writer wont to begin sentences with "and", intentionally use sentence fragments, on top of all kinds of other writing mistakes.

But a tech article this morning got it half right, and that ain't bad. The writer still conflates method and methodology here:
I've been working behind the scenes on a radically new test methodology.
Yet what follows is a discussion of why he believes some methods are better than others, and why a particular method was chosen for his review. In other words, actual methodology. This made me very happy! The article is a review of the Samsung 960 EVO SSD which is of limited interest, but within his methodological discussion is this gem
 Analyzing trace captures of live systems revealed *very* low Queue Depth (QD) under even the most demanding power-user scenarios, which means some of these more realistic values are not going to turn in the same high queue depth ‘max’ figures seen in saturation testing. I’ve looked all over, and nothing outside of benchmarks maxes out the queue. Ever. The vast majority of applications never exceed QD=1, and most are not even capable of multi-threaded disk IO. Games typically allocate a single thread for background level loads. For the vast majority of scenarios, the only way to exceed QD=1 is to have multiple applications hitting the disk at the same time, but even then it is less likely that those multiple processes will be completely saturating a read or write thread simultaneously, meaning the SSD is *still* not exceeding QD=1 most of the time.
I will admit to confirmation bias here since I've long believed QD1 to be the most important SSD metric. And until developers start paying attention to multi-threaded disk IO, QD1 will remain important in much the same way CPU frequency and memory latency are generally more important than CPU cores or memory bandwidth.

* Ironically, people who think the incorrect use of decimate should be acceptable like to defend their illogical view by accusing their opponents of employing the rather logical sounding etymological fallacy, which, Wikipedia tells me, is a species of genetic fallacy.

But these sorts of fallacies are informal ones, i.e., undeserving of sharing even part of the certitude that "fallacy" conveys in formal logic.



No comments:

Post a Comment

Note: Only a member of this blog may post a comment.