Re: [OT] WAN throughput comparisons

["David Buckley" <db@xxxxxxxxxxx>]

--- In ukha_d@xxxxxxx, "Paul Gordon" <paul_gordon@h...>
wrote:
... ballpark figure ... throughput ... ideal conditions, over
> a variety of circuits, such as:
>
> 1.5Mbps T1
> 12Mbps ATM-PVC
> 2Mbps IP-Flex
> 8Mbps Cellstream

Don't know anything about IP-Flex, but I've used all the others (or
similar to), and under "ideal conditions" (i.e. you keep the pipe
full in one direction, which for a migration is the case, as you're
copying entire volumes from sorce to destination site) you will
achieve the divide by ten byte throughput.  So a T1 will get
1.5mbit/10 = 150KByte/sec.  Assuming not NetWare with IPX!  The ATM
circuits should get better than CIR/10 due to bursting above CIR.

I assume you know that with commercial ATM systems you have a
guaranteed "bandwith", called the Comitted Information Rate
(CIR),
and the network will let you "burst" above the CIR if it has
spare
capacity in its trunks, and you have fast enough connection to the
ATM network to exceed the CIR.  So you may have a 3mbit/sec CIR and
an 8mbit/sec physical connection.  You'll always get 3mbit/sec,
guaranteed, but you may get (a lot) more, up to 8mbit/sec, some of
the time.

I've used ATM both with Bay Centillion and BCN locally and with
cisco catalyst over local fibre and WAN dark fibre (but not across a
providors cell network), and the extra packetisation steps required
above what FDDI or 1000baseF needs dont seem to make any difference.

With a migration project Cellstream might be very cool, as its gonna
happen out of hours, when the rest of the world is quiet, so there
should be plenty of burstable capacity above your CIR...