We are installing an ATM backbone connection and wondering what level
of overhead can be expected. Ive read from %10 to %50 - this will be a
LAN connection so we can assume almost no cell loss. Our provider has
said on average %12 bandwidth is overhead. It will be a Cisco->Cisco LAN
configuration. Thanks!
Stephen Balbach
VP ClarkNet
Well Stephen,
Here at ACSI, our entire national backbone is ATM, the overhead so far
seems to be about 12-14%. This is taking into account the 48/53 byte
percentage and the time to reassemble the cells into packets at the remote
end. I have run tests in our lab and we can totally saturate a DS3 and an
OC-3 link via ATM. This is in contrast to a clear channel DS-3 which
itself loses some bandwidth to conversions and overhead. I would guess
that the difference of DS-3 ATM and clear channel is around 9% of your
bandwidth but I need to run more tests in the lab to make a more educated
guess. But you don't run an ATM backbone if your just offering IP
service, we use it to offer Frame/ATM/IP services all over the same
wire. Now, packet of sonet seems the way to go for high speed IP with
little overhead, but it is only available at 0C-3 and higher. I have not
tested it yet to see the overhead or how good it works.
Anyone out there really tested the POS cards from Cisco yet?
Eric
We are installing an ATM backbone connection and wondering what level
of overhead can be expected. Ive read from %10 to %50 - this will be a
LAN connection so we can assume almost no cell loss. Our provider has
said on average %12 bandwidth is overhead. It will be a Cisco->Cisco LAN
configuration. Thanks!Stephen Balbach
VP ClarkNet
It probably depends on what you define as overhead. You might define
it as that 16 bytes of AAL5/SNAP header & trailer per IP datagram,
plus 5 bytes per cell, plus whatever trailing bytes are wasted
in the last cell.
Compare this with packet over sonet (1 byte of IFG, 4 bytes of
ppp encapsulation, and 4 bytes of CRC).
Peter Lothberg gathered some packet size distributions at various
internet routers in January, and found that using the above definitions,
atm overhead consumed 22% of bandwidth, vs. 3.1% for POS overhead.
Looking at it another way, POS can move 24% more payload than ATM
using the same packet size distribution. I've taken other snapshots
at other routers since then, and the results come very close.
/Darren