[p2p-research] What should a Community do with Profit? (was: personal server technology)

[Eugen Leitl]

On Sun, Feb 21, 2010 at 10:08:32PM +0100, M. Fioretti wrote:

> I answered to Patrick trying to stick to the original question, that
> is "personal servers". Speaking of servers means speaking of

I realize that the issue is hard to contain. I must admit I wasn't
even trying to.

> "datacenters" and of course "datacenters" may be anything from a

Technically, http://www.pogoplug.com/ is a server. Something hitherto people
have been rolling out by themselves, using the DynDNS service and home routers
which double as NAS (exporting USB hard drives and flash sticks on 
the local LAN) to make home resources available abroad. Poor upstream,
yes. Unreliable, yes. Still, a lot better than no access to home
resources at all.

Of course Pogo is still a commercial service, so what you need is a serverless/
headless P2P grid client running on your local embedded. The poor availability
and low upstream is no longer an issue if multiple copies of your content
are dynamically locally coalesced on demand. Wherever you are. As long
as you run the right client, and authenticate properly.

> closet to a big structure as I described. In any case, I sticked to
> that, ignoring on purpose the last mile and backbone connectivity
> issues. I agree with most of what you said about infrastructure, but
> that's a different issue wrt the original topic.

The topic warrants further exploration here, methinks. Anyone agrees?
 
> to point out that this is what you need to keep a *server* up and
> running. What you answered:

We're less interested in a particular piece of hardware than the
service it provides. Even few instances of low-end embeddeds on
unreliable, low-upstream domestic broadband can provide a remarkably
resilient, reliable and performant service. In theory, at least.
The proof of the pudding is in the eating.
 
> > Redundant networks (meshes qualify) degrade gracefully, so there's
> > no need to rush out for repairs.
> 
> is true, but is a whole other issue (network uptime and performance vs
> reliability and uptime of the single server). Even if I agreed that

The probability of complete outage over multiple geographic locations
and multiple providers is indistinguishable from zero.

> carrier quality systems are an overkill for _connectivity_, I would
> still want my actual server to run carrier-quality hardware.

If I'm using a service, which relies on multiple locationally and
logically diverse servers then the quality of the individual server
and its connectivity does not nearly matter that much.

Of course this is not what most people are familiar with, and 
making this foolproof zero-administration is rather nontrivial.
 
> > The only reason for a virtual server is energy efficiency,
> > professional operators and it being close to the network
> > core/backbone.
> 
> My reason for a virtual server is high energy efficiency and

Well, embeddeds do around 10-15 W these days. That's around
10-25 EUR/year, 27/7/365 operation. Virtual servers start at about 10 EUR/month,
and decent ones are 20-30 EUR/month. 

> reliability at a much, much lower price than I would have to pay to
> get the same quality of service with a home box. Again, regardless of
> the nature of the network connecting it to everything else.

Don't get me wrong, there's a reason why I'm renting a whole rack
at a major local colo provider. There's a very large difference
between 100 MBit/s Ethernet to the nearest core switch, and 
6/100 MBit DOCSIS 3.0 cable modem at the periphery.
 
> > It's mostly a marketing problem. There's plenty of VPN products
> 
> the services we're discussing here require no VPN. This said, of

VPN is a major requirement from end users, simply because they don't
want their ISP to spy on them. And VPNs are a simple way to avoid 
ISP throttling, at least for now. Won't last, of course. But the
privacy at least will remain.

> course it's marketing, but there's no demand for it yet, since most
> people are so enamored with the "cloud".

The word is a fad. Just as P2P was. In another 5-10 years there
will be new buzzwords. The problem set will be mostly static in
that time frame, however. Nothing we have today is fundamentally 
different in a 10, 15, even 20 year frame. So everybody will be
on 1 or 10 GBit/s symmetric domestic broadband. IPv6 or similar.
That's what I have on the local loop already. Easy enought to 
benchmark in the lab, or to model. 

The future's here already. Very few surprises, if you know where
to look.

-- 
Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org
______________________________________________________________
ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE