[aprssig] Tier 2 Status
Stan - N0YXV
n0yxv at gihams.org
Wed Jun 21 22:21:28 EDT 2006
Quoting Steve Dimse <steve at dimse.com>:
> No, you misunderstand the configuration of the APRS-IS. All the tier
No I do understand the configuration you're misunderstanding me. Everything you
just typed I already know, my confusion isn't in how it works but why we do what
we do.
> 2 servers are connected to the core servers in one place. Were all
> the core servers to go down, you would have a bunch of non-connected
> tier two servers. This in fact is one of my principle objections to
> tier 2. The user gets little feedback about the connectivity of the
> server they are connected to. For example, say a typical tier 2
> server has some sort of problem with its connection to tier 1. As a
> user, you are connected to a network, but have no idea your APRS
> universe is limited to those people connected to the same server. On
> the other hand, someone using the core is either connected or not, if
> you are connected to a server you are connected to everyone. If you
> are not connected, your program tells you that information.
Sounds like a problem in search of an answer instead of just a problem to be
avoided. I would _think_ that some programmer might be able to come up with
something that could forward disconnect notices. I'm _not_ a programmer so I
don't really know.
>
> Granted, it is pretty rare to have a problem with the outgoing
> connection to a tier 2 server, and therefore my second issue is the
> more common one, packets must pass through more hubs to reach users
> that do not happen to share the same hub. Image two stations each
> connected to a hub. If they use tier 2, the odds are only what, 1 out
> of 20?, that they are on the same hub. If they are not, their path is
> tier 2 -> core ->tier 2 at best. If the two tier 2 hubs are not
> connected to the same core hubs (which will be true 2 out of 3
> times), then the path is tier 2 -> core -> core ->tier 2. So the odds
> are 5% one hop, 31.7% 3 hops and 63.3% 4 hops.
>
> With core connections for the two clients, the odds are 33.3% one hop
> and 66.7% two hops.
>
> Which sounds better to you?
The difference between the two is very slight. I haven't had any trouble doing
it either way. It's so transparent that you can't tell the difference between
the two ways. I guess if you want to get down to it you might be adding a couple
of more milliseconds the long way. Your the first person I've talked to that has
ever mentioned having problems connecting to any of the servers. Could it be
that this problem is more of a localized problem in more urban areas?
> The argument for the tier 2 system is usually the core can't support
> the number of connections currently in APRS. As it stands now, that
> is a true statement, but that is simply because the core has not
> attempted to provide that much core capacity, instead the system
> keeps pace just with the number of users on the core. I believe two
> more servers in the core (which would bring the odds to 20% one hop
> and 80% two hops) would be adequate. The core servers do need to be
Great lets hop to it...what kind of time frame are you looking at to add those
next servers? Is this something we can look forward to in the coming year?
> on very fat pipes, but servers in such locations are available, every
> time I need a server location I get several good offers from people
> here on the sig.
Great I'll take an OS48 with two new 3Gig dual processor servers right away.
:-)(Sorry for the smartalc comment I just couldn't resist and it gets to my
point qwick.)
>
> Steve K4HG
>
It's obvious that you don't like the Tier 2 system but why unnecessarily load up
the current Tier 1 servers with people that don't need to be there until you get
your 4th and 5th Tier 1 servers on line? If people are connected to the Tier 1
server and don't know there is a Tier 2 to connect to that's one thing but to
encourage _more_ connections than necessary to Tier 1 is where I'm confused. Yea
it may not be pretty but we have to live with whats available now.
More information about the aprssig
mailing list