Hosting filePro applications

[Fairlight]

When asked his whereabouts on Mon, Jan 23, 2006 at 02:12:02PM -0600,
Lerebours, Jose took the fifth, drank it, and then slurred:
> 
> All of these are very valid points and worth mentioning.
> This is, of course, the way things get started; you must
> have both check lists and work on converting minuses into
> pluses.

Agreed.  You just have to make really sure that it's going to pay for
itself at a -minimum-.

> I do not see how fpTech would loose as much money as you
> make it sound.  If shared hosting or virtual server hosting
> was not lucrative, I would venture say it would not be 
> so widely offered.

I'm not sure how much you know (or anyone around here knows) about virtual
server hosting.  I've worked around it a fair amount.  Well, backtrack--I
got a client as a referral in '98 for whom I set up an ISP from the ground
up.  Negotiated all PRI and T1 costs, recommended his remote concentrator
units, administered the entire site from remote for 1.5 years after I
configured the entire thing.  I know a bit about running a service
provider, and the costs/pitfalls involved.

It may surprise you to know that most providers flat-out oversell their
bandwidth.  Having worked with the ISP, I can see why, if nobody is
capitalising on subsidiary services.  Every time you're about to break even
on, say, dialup, you'd have to lease a new PRI, get a new card, etc.  Same
thing with virtuals.  Sure, you can stick a lot of people on one machine.
All it takes is -one- bad user to screw everyone on the same machine.

Verio runs virtual BSD hosting.  I have a client on their service.  We're
currently migrating -off- of the virtual service because several times a
day the load on the machine goes so high that everything becomes -really-
unresponsive (we're talking 10 seconds to pass a simple < 20 byte TCP
packet [contents before header overhead] to a remote host, or receive same
from remote host.  At a load of 6.13, it took 3.61 seconds to get a simple
single-name 'ls' of a meager 41 files.  I've seen faster responses from a
386sx25 on a 9600bps modem.

It's offered by a lot of places that offer 99% uptime as well.  That's not
good enough.  Do the math on 99% of 365.  They allow themselves to be down
3.65 days of any given year without penalty--and call this good.  Unless
it's "five 9's" uptime, I'm not even interested.

You -really- have to look at what you get and what you can affordably give
when you go into virtual hosting and service providing.  The numbers aren't
anywhere near as attractive as you might think.

> As far as speed is concerned, filePro should not complain
> since its speed is the one thing we all brag about  8-)
> I mean, I use Godaddy.com and it moves pretty fast resolving
> GUI applications (server side scripting using CFML).

Ever logged into a system doing full index rebuilds for hours on end? :-/
I semi-regularly have to.  The system gets pretty sluggish.  It's a dual
Xeon 2.4GHz with 4GB, using hyperthreading and the load will still jack up
with pretty much nothing else running but a nascent samba that isn't in use
during that period.  Now halve that performance for only two people, and
extrapolate so on and so forth.  You can say, "Yes, but people won't
regularly rebuild all their indexes on huge files."  People do odd things.
You have to plan for them.

> Am I the only one here that would not care for some one 
> else's code?  If fpTech were to do this and I hired their
> service, I would not venture to peruse around and try to
> steal/view other applications (unless invited to by owner).

No, you're not the only one, Jose.  But solid security isn't founded on
good morals or ethics.  It's supposed to be designed to keep the people out
who have neither.  Trust is not the issue.  MAKING SURE is the issue.

> There is one thing I am certain to take with me without any
> blemish, my integrity!

Amen.

> All the points you brought up are strong issues and very vital
> for a successful launch of such service but, I think, all are
> addressable - Leave it to Ken ... before the 2nd pizza pie is
> over ... plus however many cans of coke or cups of coffee ...,
> I am sure he would have a solution in his end.

Ken doesn't strike me as the caffeine type.  Too laid-back.

> Besides, tell me if you could not come up with a security system
> to keep every one in check?  

Sure, I (or anyone that's conversant in systems and application security)
could come up with one.  Spec is the easy part.  Implementation is a bit
dicier, and we'd have to wait on fP-Tech to do that.  It has to be integral
to the engine itself, not tacked on or plugged in.

> This sort of idea deserve tossing around as alternative source of
> revenue.  Besides, it is the future!

I think it has merit as a service.  I also think it might be difficult to
provide at a price point that both floated the cost of running the service
(let alone made a profit) without making the purchase of actual licenses
for private use on your own servers look attractive in the long run.  The
only thing that might make it worthwhile in the end would be if your
service was automatically upgraded to the next release when it comes out,
and you had no upgrade cost.  Unfortunately, since there are bound to be
compatibility issues and understandably cold feet about moving to a live
implementation of a new version without months of beta and gamma testing,
you'd pretty much have to have multiple versions all available at once, and
running modules of themselves at once.  That could be messy alone.  Then,
when someone gets off the pot and actually decides they want to switch,
someone has to make that happen--enter your "upgrade cost".

I swear I'm not trying to shut you down, man.  Keep thinking it over and
countering me on this.  I haven't done research with modern numbers on the
connectivity -or- the hardware, as I haven't been asked to do so for a
client in a while.  I just know the nature of the beast, and have a good
idea of how the numbers tend to turn out in reality as opposed to theory
when you go into something like this kind of service.

Your key is -volume-.  Without volume, you have nothing except red ink.
It's the same thing as selling the software itself.  If they could do that,
we wouldn't be contemplating alternate revenue sources for them, right?
They don't have enough recognition in the marketplace to be able to get the
volume to support this, even if 'x' amount of critical mass would float
the boat.  And at that point, they'd be better off selling the licenses
in the first place as that has far less operating costs.  They have to
pay for development they're doing in either case.  In the case of just
selling the software licenses, they incur no extra charges.  In the case of
providing a service as discussed, they incur a lot of expense, at a loss of
net financial potential, with added legal exposure.  Not a good strategic
move. 

Call me crazy, but I say, much as I like the -theory-, that the bird will
not actually fly.

mark->