I see how that could present a problem ... followed by the next obvious
question of why aren't you sleeping at 4:00AM?

Seriously though, you might find a group locally that you could meet
face-to-face because of the magnitude of your questions. That said, let
me take a stab at some of the easier ones.

Hosting providers come in several flavours.
- people who let you put your server in their datacentre and you manage
it remotely (expensive);
- people who install your application on their server and manage it for
you (very expensive);
- people who install packages from a menu and you and they use varying
technologies to limit your view on the server and you are actually
sharing the server with many other customers but you can't see/interact
with them. (Inexpensive) Obviously in this model you have no control
over the machine itself - you simply call their support desk and open a
ticket.

The third option can be viable but you need to find a provider that
offers Asterisk as part of their service because you are not at liberty
to install new applications that aren't on their supported list.

In the third model it also means you wouldn't need to provide them with
software and teach them how to install it because they do everything
and  instruct you how to interact with them.

In the first model, it's your box and if you need to get on a plane ...
then you need to get on a plane ;-)

In the second model they take care of everything for you but it is
expensive because they need to come up to speed on your application,
learn how it works, learn how to integrate it into their Managed
Operations environment, etc.

So for your purposes I'm assuming you are probably looking at options 1
or 3 depending on volumes and remote access abilities.

OK, remote access.
Unix (meaning Unix, Linix, AIX, DG/UX, HPUX, etc. -- all the *ix's) come
with significantly more "stuff" in the box than DOS/Windows ever did.
You can do a lot of functional work without having to buy additional
software. Remote access is one of those.

Unix servers support a protocol called Secure Shell (SSH) which encrypts
the entire communication stream. A managed provider should offer you SSH
access to a shell prompt or "shell account" (Unix terminology for
command line). From there you can manage most of the box as Unix's have
standardized on using plain text files for configuration meaning that a
simple text editor (which comes with Unix) can be used to change all
your configs. (AIX excepted as they use something called ODM which is
akin to Windows registry - blah).

OK, command lines good, but what about a gui?
Your in luck. Contrary to Microsoft Windows, the windowing system for
Unix - Xwindows - separates the actual pictures on your screen,
keyboard and mouse from the application doing the work by a TCP/IP
connection. This means the GUI can be drawn on one machine while the
application runs on another far away. SSH from your shell account can
be used for this so you can start the graphical environment remotely to
manage the box.

One of your other questions was relating to the box' ability to keep up
with demand. Now there is no free lunch here. VoIP is great or we all
wouldn't be here, but from a quality consistency standpoint it is
always a compromise. The key is at what point on the curve do you place
yourself. Generally speaking a VoIP box needs to be over provisioned to
handle it's task and shouldn't be bogged down with additional
applications because of they way time slices are allocated.

Lets look at old-style telephony. Time Division Multiplexing (TDM) takes
a stream (lets call it one channel - 64K) and reserves that space by
sending constant traffic between both end-points. When the channel is
not used it sends filler, when the channel is used, it replaces the
filler with the content of the audio call.

Not real efficient because it means you transport full bandwidth ALL THE
TIME. The up side, is that because that bandwidth is both physically and
logically reserved, you can never get "bumped" because there were not
available time cycles for the cpu (cause it was already generating
filler) or in the pipe (because it was already consumed). But ... the
voice quality is stellar.

Economies being what they are people want to move from TDM to
packet-switched (VoIP). This means just like every other application on
your PC, packets get sent only when necessary. The upside is obvious on
communication costs. The downside is that at the time you need to send
something, the CPU may be too busy to fulfill your need and/or the pipe
may be too full to transport the data.

This works fine for data. Your web browsing can be "slow" but it still
works. Voice packets however, have a "time" dimension that must be
adhered to. If the packets don't arrive within the alloted time to be
reconstructed into voice they may as well have not been delivered at
all because they become useless.

Hence, VoIP servers need to be working well below their capacity to
ensure they have the muscle to deal with the call at the time it
happens. That said, there is a little latitude - remember the "place on
the curve part"? There are technologies used to ease the timing problem
in the likes of jitter control and echo cancellation.  And they work
well, but remember that they are both means of correcting errors when
they have occurred. It is much better to have excess capacity in the
servers so the problems don't occur and need to be corrected in the
first place.

OK. I think I'm done for now. Somebody else can take over (or refute my
comments as there is never a shortage of differing opinions).


dbc.
--
David Cook


Quoting Chriswlan2 <[EMAIL PROTECTED]>:
> ...
> Only trouble: I'm just outside Calgary...
> Sigh...
> :-)

Reply via email to