LinuxSA Mailing list archives

  From: Richard Russell <richardrussell@mail.com>
  To  : linuxsa@linuxsa.org.au
  Date: Tue, 11 Apr 2000 22:47:31 +0930

An Idea...

OK folks, here it is... This is the idea that I have been
fantasis^H^H^H^H^H^H^H^Hthinking about for the past year or more... I don't
know if this is possible, or if there would actually be a market that exists
for the service I am thinking of, but suspect that there would be, and this
market would be largely untapped.

I must preface this by saying that I'm not exactly a guru sysadmin, in fact,
I'm not much of a sysadmin to speak of at all... I also don't have a great
deal of expertise in the technologies mentioned, otherwise would probably
start this project myself... I know enough to be reasonably sure of the
technical plausibility of this idea, but not of the practical
remifications... Anyway, if this all appears completely infeasable, please
tell me, and I'll either explain that I've allready thought about that
problem, and here's a solution, or I'll say "Oh... bugger", and return to my
hole...

The Market:
Any business, organisation or group who wants a basic, simple server. They
need things like SMB, SMTP, POP3, IMAP, HTTP, Proxy/Cache, Firewall, DHCP,
DNS, etc, etc, etc -- all things that most servers can do (NT, NetWare,
Unices, Linux). However, they have neither the time nor the expertise to
maintain even a server of this simplicity (and paying for the service of
configuring and maintaining their server is expensive with current
mechanisms -- even remote administration). They want their server to be very
reliable, UPS'd and backed up, and don't have any major special cases (eg,
they don't need to use Linux for development work, and don't need Exchange
server, and so on...) I imagine that there would be a large number of such
organisations out there... although I may be wrong -- please feel free to
contradict me... Actually, this idea would also extend to any large
organisation which would like to replace their departmental NT servers with
administered Linux boxes...

The Service:
Provide a Linux server that will do all the things that Linux is known to do
well, and maintain it remotely (yes, this isn't really new, but the exciting
bit is in the next section -- the bit that makes this a cheap and economical
service to provide). As far as the client is concerned, they don't care that
it's Linux, and won't even need root access. All you need to provide them
with is the services they require, and an internet connection (this could be
dynamic IP, but static permanant is better). The Client simply needs a means
to shut the server down (in case they need to move it, or something), they
need access to the tape drive, so they can swap the weekly/nightly backup
tapes, and they need a network to attach it to. As well as this, the client
doesn't really want to invest in Computer hardware that will go obsolete
within a few years, so the system is provided to them as a service, with a
monthly fee for providing the hardware, network connection and the
administration (basically like outsourcing for small business)

The Exciting Bit:
If you can get a large-ish number of such clients, all with pretty much
similar needs, then you can create a database with all their details in it,
and administer them in a semi-automatic fashion. for eg: a new kernel is
released, so you update your database (which includes the .debs or RPMs that
each machine uses) with that kernel. The next time each client connects to
you, they download an automatically generated list of debs/rpms to install,
then proceed to download them, and once they are all downloaded (and at an
appropriate time -- 3am), install them. If the differences between each
client are small, then each client will effectively have the same set of
changes, and the amount of administration needed per machine is small -- you
just interact with a database, and administer large numbers of machines at
once.

Where individual differences exist (particularly in password and group
files, as well as host-specific info, such as hostname, IP number, PPP (or
whatever) connection, and so on) it is relatively easy to see how you could
have these bits of information stored in the database somewhere, and
generate a .deb or .rpm containing the configuration files from the bits,
PHP style. If some level of control is desired or needed by the client
(partcularly for groups and users and dns entries), then some simple CGI
scripting should allow the client to add new users, new groups, and put
users in groups as they wish, it's just that these changes would only be
stored locally, and wouldn't be known about by the server, but as long as
they are backed up, this wouldn't be an issue...

Scenarios:
1: Bob's Accounting Firm uses this system, mainly using Samba for storing
all their documents, Apache for an intranet, and a few other generic
services, such as email and Squid proxy & caching. A cleaner accidentally
spills a bottle of alcohol-based cleaner on the server, then pulls the plug,
causing a spark which sets the server on fire. Everything appears lost, but
with the database and the backup tape, the providers of the service are able
to replace the server in less than a day (assuming hardware can be found),
and work returns to normal (apart from the smell...)

2: The First Professional National Real Estate chain of real estate agents
has over 200 offices around Australia, each using this system. A serious
exploit is found in sendmail that compromises all their confidential
business data. Assuming a .deb or .rpm has been released with the fix, the
fix is set for deployment in a few clicks of an administrator's mouse, and
that night, each of the 200 machines fix themselves. What would have been
200*10minutes + travelling time, turns into 10 minutes.

3: You are the provider of the system... OpenExchange 1.0 is released (a
hypothetical Open Source, Linux Based MS Echange equivalent), which allows
users to use all the facilities of MS Outlook. After some preliminary
testing on a group of test machines, clients are contacted, asking if they
would be willing to pay a little extra, and have full use of Outlook. Some
say yes, so a field is added to the database, and anyone who wants this
service has it installed within a week -- total cost of installing new
service for the service provider = cost of testing + cost of adding
funtionality to database (pretty minimal if it is planned for) + time for
sending bulk email sales speil to all clients + a few minutes per client who
wants the service to turn it on for them....Total revenue from providing new
service = number of clients who want service * price per month charged for
service... It's not hard to see how this provides a pretty constant cost,
with a growing revenue...

Hopefully, I have explained this well enough... if you have any doubts that
this would work (and sell), please tell me, as I would be interested to
know. Also, if anyone is interested in actually starting such a project (or
is aware of one already existing), I'd be pretty keen to know about it :)

Hope you enjoyed my flight of fancy... :)

rr

-- 
LinuxSA WWW: http://www.linuxsa.org.au/  IRC: #linuxsa on irc.linux.org.au
To unsubscribe from the LinuxSA list:
  mail linuxsa-request@linuxsa.org.au with "unsubscribe" as the subject