Re: [Mratc-developers] Define some ATC specifications

[Michael R. <ht...@us...>]

----- Original Message -----
From: "Aquila Deus" <aqu...@ya...>

> If we plan around handling US air traffic volume, we should be
> able to handle volumes for any part of the world.  For the US,
I
> think a good maximum for the number of flights in the air at
one
> time is about 223,440.

How about multiply the number by 10 or 100?

Why multiply by 10 or 100?  We can build a scalable system, but
initially, we just need to handle what's there.

> I have one good news and one bad news:
>
> The good news is I found a JVM optimized for extremely
large-scale
> applications: BEA's JRockit. And it supports lightweight
threads (called "thin
> threads").
>
> The bad news is the lastest version of JRockit seems to have
removed the
> lightweight threads :D
>
> I don't think it's good to use native threads or processes to
emulate millions
> of flights. Maybe we should really consider erlang seriously
(but then we will
> have to use emacs). Do you have a better solution?

I was thinking of approaching it with Perl and MySQL; you are
invited to put together your best approach at the problem
(unless you want to come along with me and Perl); and we will
compare the two.  I think the scale involved here would benefit
from a database, with its indexing capabilities.  If we both
develop separate systems, we need to conform to some basic
definitions (like having the same lookahead interval size and
number and separating general flight direction by altitude),
though, so we can compare apples to apples.

Threads can be done in Perl, if you are interested.  Yes,
223,440 threads sound like a lot.  I've been thinking iteration
all along, rather than threads, so that number of flights does
not discourage me.  However, we might get an ATC system put
together that takes an hour to iterate through 223,440 flights.
If that is the case, we can divide and conquer.

--Mike