[HN Gopher] Two workers are quadratically better than one (2020)
___________________________________________________________________
Two workers are quadratically better than one (2020)
Author : Tomte
Score : 50 points
Date : 2022-05-02 18:35 UTC (4 hours ago)
(HTM) web link (www.hillelwayne.com)
(TXT) w3m dump (www.hillelwayne.com)
| marcosdumay wrote:
| As a rule, if you have more than 50% of utilization, latency
| grows very quickly. That is true with any numbers.
| williamkuszmaul wrote:
| I'm not sure why they claim that the total time grows
| quadratically.
|
| If tasks arrive arrive randomly at the same average rate as they
| can be processed, then the amount of time that the nth task will
| have to wait is proportional to sqrt(n) in expectation. So one
| would expect a total waiting time of n^1.5, which incidentally
| fits much better to their plotted curve than n^2 does.
| travisgriggs wrote:
| It's funny. I saw the headline, and my immediate association was
| "employees" rather than computation tasks. And I was thinking
| "not even!" It's been my experience that scaling up employee
| count, has diminishing returns, with even some net negatives with
| some adds.
|
| I'm not sure I buy the quadratic gains, but the general gst is
| that if done correctly, two processes/tasks/workers/etc is better
| than one.
|
| So why is this? Why does it work in computing, but not with
| people? Is it the nature of the tasks (one is problem solving,
| the other is grinding out computations?). Is it differences in
| configuration?
| rootusrootus wrote:
| Discussion from 2020:
| https://news.ycombinator.com/item?id=25001875
| civilized wrote:
| Two workers are infinitely better than one for latency, if you
| have slightly more work than one worker can do
___________________________________________________________________
(page generated 2022-05-02 23:00 UTC)