Which is objectively the best one?

>>57593919
depends

>>57593919
it depends on what you're trying to implement. If I were Netflix, nothing would be worse than decentralized or distributed

What are you trying to do?

>>57593919

Distributed. Name one actually important thing that requires other model: you can't.

>>57593919
all 3 models are false

here is a visualization of 1000 friends in a social network. which model is it? i have no fucking clue. this is because 1000 friends have 1000^2 = 1,000,000 degrees of freedom to form relationships, compared to the tens of relationships in these goofy model drawings

with just 1,000 users it's difficult to tell what we're looking at. facebook has over a billion users. these models are just toys that are radically detached from the realities of social structures

t. graph theoretician

File: 1000friends.png (678KB, 1148x734px) Image search: [Google]

678KB, 1148x734px

>>57593992
oh my god i forgot the fucking image fuck

here you go

>>57593974
Performance, especially latency

>>57593924
this. every other answer is retarded. stop being such fucking retarded zealots.

>>57593997
that's decentralised, brainlet

>>57593992
>all 3 models are false
Lol, are you so stupid that you don't even understand what false means?
How are these 'false'? They are fucking models.
And where did OP imply that they have anything to do with social structures?
They are computing related, as you can see by their subtitles.
I seriously hope you are drunk right now, or I will have to lose all my respect towards 'graph theoreticians'.

Anyway OP, if all hops take the same resources, then centralized is best.
Fewest hops between any 2 nodes on average.
Irl this isn't the case, so decentralization is your best bet.
Complete distribution is nice as fuck from other perspectives. Especially if it is likely for nodes to stop functioning.
But obviously the act of distributing will become very hard, especially with a dynamic topology.

>>57594087
It isn't though.
But that doesn't matter, because OP is talking about something different.

>>57593939
Don't cdns make Netflix and other similar services distributed in a way

>>57595020
Yes they do.
I also doubt that they have one single facility where things get served from.
A decentralized model is more likely.
Their back-end is decentralized anyway, like any really big website.

>>57593992
>>57593997
that looks decentralized to me

File: monk.png (582KB, 566x789px) Image search: [Google]

582KB, 566x789px

Distributed has most built in redundancy by virtue of its nature, so it's what I tend to opt for.

>>57593997
>A few vig hub nodes with lots of connections and then outer nodes with fewer.

It's decentralized mate

>>57593919
centralized for complete control
decentralized for maximum efficiency
distributed for maximum availability

>>57593919
COMPLETE NETWORK.
EVERY NODE CONNECTS TO EVERY NODE.
FINAL DESTINATION.

>>57594087
>>57595577
>>57596183
That picture does not represent decentralized.
>>A few vig hub nodes with lots of connections and then outer nodes with fewer.
Look at decentralized again
>A few big hub nodes with lots of connections and then outer nodes with ONE connection

But this guy got it: >>57594914

Spotify uses distributed p2p streaming.
Skype was the same(called Big nodes) until Microsoft came in and put everything on their data centers, after the disaster Windows update.
Napster(yes, it still exists), uses o combination of Client Server + p2p. First, the server finds the peer that has the required file, and then it establishes a connection between you and that peer. At least that's how it worked back in the day.
See, it all depends. There is no definitive answer.

>>57597870
Cont.
Spotify can quickly switch between p2p and client server, depending on the situation and streaming quality.

>>57597887
>>57597870
/g/ doesn't get excited about informed, nuanced answers.

The first one on the left is the best one, because it helps prevent terrorism and ensures that everyone participating can feel safe.

>>57593997
looks decetralized, buddy

>>57598785
Nice try NSA

>>57593919

Easy:

You don't have to worry about system failure or you have a clear distinction between sender(s) and reciever(s)?
Pick nr. 1

You have a general concern about "intranet vs. internet", ie. a lot of communication is intranet (think universities, where most mails are within one "group"), or maybe you want applications that are only available via intranet (security concerns)?
Pick nr. 2

Highest redundancy (muh ARPA), the most simple structure or you're not sure about which nodes will communicate with each other eventually?
Pick nr. 3

>>57593919
token ring

I worked for the logistics institute that planned the new postal system for both the reunited Germany and for Austria, and we've used the middle one.

>>57600242

Rolf: "Fünf ist Trump".