Mailing List Archive

My recollection is that ACTS is free for use by public universities, but
that the footprint doesn't reach very far north, and that the latency is
high enough that not very many schools are actually using it.

-Bill

> High orbit, geosyncronous sattelites do not stand much of a chance against
> land lines as the latency on the links is quite high. [...]

Long latency is not automatically bad. It is bad for interactive traffic,
but if the bandwidth is high enough to reduce congestion to zero, a large
latency doesn't hurt bulky transfers at all. Netnews, for example, could
be distributed via satellite without hurting anybody's lookers or feelers.

You can always have compartively small land lines to handle burst
traffic. And the bulk of the data [lower priority] over the satellite links.

[Similar to the way Oracle handles interactive video over cable.]

-Deepak.

On Sun, 1 Jun 1997, Paul A Vixie wrote:

> > High orbit, geosyncronous sattelites do not stand much of a chance against
> > land lines as the latency on the links is quite high. [...]
>
> Long latency is not automatically bad. It is bad for interactive traffic,
> but if the bandwidth is high enough to reduce congestion to zero, a large
> latency doesn't hurt bulky transfers at all. Netnews, for example, could
> be distributed via satellite without hurting anybody's lookers or feelers.
>

According to Paul A Vixie:
>
> > High orbit, geosyncronous sattelites do not stand much of a chance against
> > land lines as the latency on the links is quite high. [...]
>
> Long latency is not automatically bad. It is bad for interactive traffic,
> but if the bandwidth is high enough to reduce congestion to zero, a large
> latency doesn't hurt bulky transfers at all. Netnews, for example, could
> be distributed via satellite without hurting anybody's lookers or feelers.

And with the push to use multicast across such high latency links, you
can get better dessimination of some types of data which involve many
sites within a satellite's foot print (e.g. nntp, ftp mirroring, db
syncronization).

There also is a new proposed spec for doing FTP using multicasting
(MFTP - multicast FTP) which has been tested using some INTELSAT birds
with good results. MFTP has also the characteristic of reduced
sensitivity to long latency, but also to the bit error rate (BER) which
is also a problem when using satellites.

--curtis
--
Curtis Generous generous@uucom.com Phone: (703) 461-1350
UUcom Inc., Suite 250, 4875 Eisenhower Avenue, Alexandria, VA 22304-0797

>
> I have been following this thread with some interest. A few months ago a
> friend of mine from Fore Systems told me about the 'ACTS' project. NASA and
> Ohio-State and quite a few others were doing research into the Gigabit
> Satelite Network.

ACTS has been used for a bunch of stuff, but you're probably thinking of the
high data rate experiments. The home page for ACTS HDR is
http://mrpink.lerc.nasa.gov/gsnhome.html

> There was a test between Sony in Japan and I 'BELIEVE'
> JPL in Pasadena of hight def video transmissions over an OC-3 circuit.

You'll find this on on the Experiments page, number 101.

> Anybody heard anything about this? If enough birds like ACTS were
> sent up, this could be a boon for ISP's maybe.
>

Well, maybe. ACTS is a GEO satellite, and interactive use is painful. On the
other hand, it's enabled people to do interesting things with bulk data
transfer, and it was probably the first system that got you OC-12 across
any significant distance. There's an IETF WG working on the issues posed by
IP over satellites, whose home page is http://www.isr.umd.edu/CSHCN/Links/IPoS/

...arun
---------------------------------------------------------------------------
Arun Welch 5000 Britton Rd
Lead Engineer, Internet R&D Hilliard, OH 43026
CompuServe Network Services awelch@compuserve.net

Perhaps this is finally a case for those little TOS bits in an IP header?
Interactive stuff like Telnet traffic could go through a slow land line,
whereas bulk traffic could go via satellite.

I'm not sure how many TCP/IP stacks still set those bits; it may be
necessary to have a router manipulate the bits after examining the port
numbers of a connection.

FYI, Planet Connect (and others) has been offering Usenet feeds via 18"
dish for quite some time now.

Stephen


At 14:33 01-06-97 -0700, Paul A Vixie wrote:
>> High orbit, geosyncronous sattelites do not stand much of a chance against
>> land lines as the latency on the links is quite high. [...]
>
>Long latency is not automatically bad. It is bad for interactive traffic,
>but if the bandwidth is high enough to reduce congestion to zero, a large
>latency doesn't hurt bulky transfers at all. Netnews, for example, could
>be distributed via satellite without hurting anybody's lookers or feelers.

> Hi,
>
> Stephen Sprunk wrote:
>
> > I'm not sure how many TCP/IP stacks still set those bits; it may be
> > necessary to have a router manipulate the bits after examining the port
> > numbers of a connection.
>
> Looks like a good place to use access lists that choose packets based on
> tcp, udp, and port number. If I could get a 45 Mbps circuit over high-latency
> geosatellite, one-way, I could use it for http and "push" traffic.
> The requests, outbound stuff, and interactive traffic would fit into a T1
> under the ocean.

if you're interested in that, cisco's new CoS stuff uses precedence
(extended access lists, mac address, ports numbers) to do just such a
thing. it would be most useful if isp's could have concensus on the
precedence "classes" to work across the multiprovider internet (may
not be practical though).

-brett