News for hackers. Stuff that matters.
Re:Counter-intuitive results in operations research by cortesoft on Saturday January 29, 2022 @02:19PM (#30129472)
Bufferbloat is something that seems counterintuitive at first glance. Why would adding a bigger buffer hurt performance?
Re:Counter-intuitive results in operations research by digikata on Saturday January 29, 2022 @03:46PM (#30130244)
And this would have logistical parallels with the just-in-time inventory strategy. which imho has been a maligned in that it does trade off other risks, but it also makes for better supply chain responsiveness (if the buffers are selected correctly on the right items).

Selecting appropriate buffer sizes is critical - a good set of comparisons there is what buffer would one want for streaming a movie vs a buffer for streaming a live video meeting connection.

Re:Counter-intuitive results in operations research by tome on Saturday January 29, 2022 @04:32PM (#30130731)
JIT has been maligned by those who don't realise that in the original Toyota conception one's suppliers should be right there with you on site. Supply chains stretching across the oceans just aren't a thing in properly done JIT.
Re:Counter-intuitive results in operations research by kqr on Sunday January 30, 2022 @01:04AM (#30134525)
Well... part of the purpose of just in time is to surface problems quickly. Transient problems in an upstream process can be papered over by inventory on the receiving end, but when you remove that inventory everyone gets a clearer view of where the problem is, and everyone can do a better job of fixing it.

So just in time over the oceans works precisely as intended: it gives us information immediately when there is a problem. Whether or not it's within any manufacturers power to fix those problems is a separate issue.

(And I'm going to guess that there's no easy way to fix those problems, given that Toyota has opted for more inventory to hide those sorts of transient issues when it comes to certain parts.)

Re:Counter-intuitive results in operations research by dtaht on Sunday January 30, 2022 @01:29PM (#30139694)
Queue theory in general applies not just to supply chains, or bufferbloat, but explains a lot about how to do JIT and other forms of "traffic scheduling".

Not enough people have read Len Kleinrock's wonderful books on this subject, and they are long out of print, and available only on archive.org.

I wish it were taught more. His Queue Theory volume 2 is one of the most well-thumbed books on my bookshelf.

Re:Counter-intuitive results in operations research by pphysch on Saturday January 29, 2022 @03:33PM (#30130112)
Can analysis of bufferbloat help us understand the current state of the US/global economy?

(Money = buffered economic value; too much liquidity = bufferbloat?)

Re:Counter-intuitive results in operations research by neffy on Saturday January 29, 2022 @05:03PM (#30131003)
It's not a bad idea per se, but no.

The actual issue is too much debt, alongside too concentrated money. The issue at some level is some buffers are far too big, and 70% of the US population doesn't have any buffers at all.

Re:Counter-intuitive results in operations research by pphysch on Saturday January 29, 2022 @06:20PM (#30131649)
I agree that debt is the underlying issue, but if we consider that the Fed's only solution to the "network issues" is to "increase the buffer sizes across the backbone" then perhaps "bufferbloat" is contributing to some of the supply chain issues.
Re:Counter-intuitive results in operations research by throwaway29303 on Saturday January 29, 2022 @04:46PM (#30130877)
Re:Counter-intuitive results in operations research by Aachen on Saturday January 29, 2022 @03:31PM (#30130091)
Post an answer :)

Here it will get lost (semi-ephemerality is imo one of the disadvantages of HN), on stackexchange people will find it.

Re:Counter-intuitive results in operations research by cortesoft on Saturday January 29, 2022 @04:41PM (#30130809)
I don't really feel like making a stack exchange account
Re:Counter-intuitive results in operations research by lucb1e on Saturday January 29, 2022 @07:30PM (#30132210)
That's one of the best things about stack exchange: it's not a walled garden. You can post answers without creating an account and the added content is then available under creative commons.