Gary Jackson wrote:
>
> One thing to be aware of is that there needs to be a clarification of terms.
> There is point to point T1, which is an empty pipe between two locations and
> Frame Relay T1, which is a service where packets go into the "cloud" and pop
> out at the Frame Relay service at "the other end".  As I recall, both ends
> need to be on Frame Relay.

This should probably be taken offline, but just for kicks:

A basic elemental telephone 'frame' is 64kbps, or a DS0.  ISDN started
with providing you with 2 DS0s or 128kbps.  Depending on the framing
type and the telco switch, this can be two B-channels of 56kbps and
one D-channel for signalling and control of 8kbps.  This comes from
"bit robbing" every 8th bit (1/8th of 64kbps is 8kbps, leaving 56kbps)
of the stream.  Modern data-oriented ISDN allows you to skip the
D-channel an dedicates the entire 64kbps to data.  For a single
connection, you can go 56kbps or 64kbps, using PPP encapsulation,
usually at the same rates as analog dialups.  In fact, many access
servers can auto-detect ISDN vs analog and serve both purposes.  If
you want full ISDN at 128kbps, you have to use multilink PPP, where
the two channels are bound together.

Now, take 24 DS0s plus an 8kbps signal/framing overhead and you have
a T1 at 1.544mbps.  In PBX or telco terms this is a PRI (the regular
2-channel ISDN is a BRI) -- Primary vs Basic Rate Interface.  The T1
itself can be framed various ways -- when channelized for PRI use it
is often 23 DS0s plus 1 DS0 for control.  When the framing is more
serialized (like Extended Superblock Framing, or ESF) you can use
all 24 DS0s for the 1.544mbps on a leased line.

Frame Relay breaks the rules a bit.  You may get a T1, or part of a
T1 (fractional T1) at the interface level, but that isn't necessarily
your bandwidth.  Frame Relay has a CIR (Committed Information Rate)
which the telco guarantees that belongs to you.  There is also the
interface speed itself which is the BIR (Burst Information Rate).
The BIR is almost always greater than the CIR (and there are some
unusual cases of CIR being zero).  In this case, you are guaranteed
the CIR, but if nobody else is using the bandwidth, you can speed
things up to your BIR -- but those packets get marked as "discard
eligible".  If the Frame "cloud" becomes congested, discard eligible
packets are dropped first.  There is also some primitive flow control
in this case in the form of FECN/BECN (Front-End/Back-End Congestion
Notification) which may or may not be honored by your router/telco.

In theory, the telco (I keep saying that, but perhaps I should just
say "ISP") will never oversubscribe a line beyond the sum of the CIRs
in that part of the cloud.  But that is just theory :-)

> If you want to use DSL (which I use here in my office) you have to use an
> ISP that provides DSL services.

DSL (particularly ADSL) and cable modems go even further -- they will
very often oversubscribe their circuits.  DSLs in a given area tend
to be grouped by a high-speed line to the telco, and the sum of DSL
bandwidth almost always exceeds the trunk/uplink; likewise with cable.
Cable is worse as it is usually shared media (sniff your neighbors)
and some implementations of DSL are likewise not 'dedicated' like
other services.

T3 is just a bunch of T1s, OC-3/SONET are another form of T3.  The
bandwidth continues to increase with OC-12, OC-48, etc.

Hope that at least clarifies Frame Relay and DSL a bit.

Jeff Kell <[log in to unmask]>