I really think decentralisation is the future of this stuff, but it's a way off yet.

Generally though: w3rd bro (hug)

What was that decentralised messenger app from yonks ago?

Hmm, there's Jabber which is a protocol (now used by Google Talk and hangouts and all that). You mean that?

IRC ?


Not fully decentralised, but segmented and stuff.

Nah, I'm sure there was a thing where.... Maybe I imagined it :$

10k, really?

 

Not that I fully understand how your theory works, but 10k isn't going to get you an inch off the ground if you want enterprise class gear.

 

Unless you don't believe in speed, reliability and redundancy.

 

 

the best forum software on the planet is free

and I am glad it is, hopefully that might encourage more people to use it
then I wouldn't end up on shitty hard-to-use forums
where I just give up more often than not

You're not thinking clearly.

How's that then? I just don't 'get' how you think you're going to host anything mildly successful for 5 or 10k. I've installed network cards that cost that much.

Well you're either exagerrating or were ripped off for that, but that's entirely beside the point.

For £10k a year I could have between 100 and 750 servers, running on industry standard equipment in world class datacenters.

And it's unlikely to actually need anywhere close to those numbers. Even considering redundant clusters and geolocation and so on.

http://h10010.www1.hp.com/wwpc/uk/en/sm/WF06b/3709945-3709945-4202225-3722810-3722810-3794423-3794432.html

 

I don't make things up :) And we always put 2 in for redundancy!

 

Here's some CoLo prices, too:

 

http://www.redstation.com/colocation.html

 

So, figure at least £1400pcm for a single rack in a single data centre. There's probably loads of extras you'll need too.

 

But whatever, Pete - we clearly disagree on a number of things such as "industry standard" and "world class".

He won't actually be buying the servers, just renting.

Well, firstly, I don't believe you can run something that big on rented hardware. Facebook has over 800 million users.

 

Let's say you store just 500 kilobytes of data per user. That's

 

800,000,000 * 500 = 400,000,000,000 kilobytes.

 

400,000,000,000 kb = ~372 Terabytes of storage alone.

 

Secondly, even if it as all rented and managed by someone else - you're still going to pay for it!

 

I just think Pete's got his head in the clouds on this one.

Are you forgetting the bit where there's the word decentralised?

The point being to keep any centralized aspect to an absolute minimum, and not be storing anywhere close to 500kb per user, probably not even close to 1kb, but for arguments sake, lets say 5kb.

1 billion users x 5 kb ~= 5 terabytes, or 16x320gb drives - not really that much by today's standards.

Certainly within the ~£1k a month budget I plucked out of the air earlier.

Okay, I think the whole concept is a bit of a joke (So, people can only access my photo albums when I'm online, then?) but even so.

How much do you think even a little starter SAN is going to cost? Here's a nice cheap starter:

http://www.systemactive.co.uk/Buynow-id-17287-HP-LEFTHAND-P4300-G2-7-2TB-SAS-STARTER-SAN-SOLUTION-IN.html?utm_source=google&utm_medium=base&utm_campaign=products

That will just accommodate what you've just mentioned.

Can you not upload photos to your fictional social network? (It was a social network, wasn't it? I'll admit I haven't been paying attention) Cos just one photo's going to blow your 5kb per user target out of the window.

I have a feeling it's going to involve some kind of peer - peer stuff where bits of data are stored on peoples machines or some such. Even ignoring the security issues there, I still think it would be horrendously unreliable and I don't know what happens with mobile stuff.

 

I just Pete has massively underestimated what it genuinely costs to run something on the scale of Facebook. I don't doubt one could write a better platform and so on, but I think they'll all get cold sharp shocks when they try and scale it into the millions of users.

Pete, seriously...?

Ok, if you go to the first post of this thread, you can see some pictures.
You can see them even when my computer is turned off - magical!


Why on earth would I buy an overpriced black box and create a single point of failure!? That's crazy. If I needed to go down the path of a centralised database, I'd use something built from the ground up to be scalable and redundant.

That's even assuming such a route is necessary - I haven't sat down and worked out specifics of what would be required, but I do think it's something that could potentially be avoidable.

As I said, the earlier £5k/£10k came from some very very rough estimates of what might be involved, simply as a theoretical example of an amount I personally would be happy to "throw away" on costs, if it would allow a real alternative to Facebook (even one only used by the people I directly care about / am connected to).

I wasn't saying "I'm going to create a billion user service for almost nothing", but I also don't agree you need to fling gold bars at HP and IBM and friends in order to run a successful service.


Your general approach here seems to be not thinking fully and/or jumping to incorrect conclusions, based on a non-existent scenario, and I'm not entirely sure why, but I'm not in the mood to keep trying to explain stuff either. :/