HP3000-L Archives

September 2004, Week 1

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Jeff Kell <[log in to unmask]>
Reply To:
Jeff Kell <[log in to unmask]>
Date:
Thu, 2 Sep 2004 02:47:06 -0400
Content-Type:
text/plain
Parts/Attachments:
text/plain (71 lines)
Tom Hula wrote:

>We recently switched from a DSL Internet line with 384Kbs
>maximum download to a cable Internet line with a maximum
>of 4 Mbs. Not seeing anywhere near 4 Mbps, the next step
>was to try individual lines hooked directly to the cable modem.
>
>
I've seen many suggestions, just wanted to throw in my $0.02 worth.

You mentioned DTCs, which as far as I know are ONLY available
10Mb/half-duplex.  Your 3000, unless it is an A- or N-class or has been
updated with a 100BT card, is also 10Mb/half-duplex.  So whatever
device(s) these are connected to should be 10Mb/half-duplex.  You
mentioned hubs.  A hub will be half-duplex, even if it is a "10/100
hub".  This latter flavor of 10/100 hub, and in some cases (such as Bay
Networks) they had some 10/100
"switches" that weren't really 10/100 at all.  It had two backplanes,
one at 10 and one at 100 with no bridge between (requiring an uplink
from each to get it to work).

If you have a 10/100 "hub" it will not only be half-duplex, if any port
of the hub is running at 10Mb, it forces the entire hub to 10Mb unless
it has buffering and bridging (very few if any hubs to store-and-forward
which would be required).

You distances sounded OK for Cat5 or 5e.  You get 100m (usually
recommended 90m wall jack to patch panel, leaving 5m each jumpers for
connecting the end devices).

Switches are a definite improvement.  Most switches these days are
store-and-forward, which introduces a small amount of latency but
isolates errors and if full duplex (at least switch-to-switch) avoids
collisions.  Some switches can still be around (haven't been shopping
around since we have essentially gone with Cisco) that are NOT
store-and-forward.  These include:

* Cut-through (or "fast") switching -- as soon as the device recognizes
the destination MAC it starts forwarding the packet.  This minimizes the
latency of the switch, but can also propagate errors (CRC, checksum,
collisions).

* Fragment-free switching (may be known by other names) -- the device
waits until the minimum frame size has been accepted (64 bytes) before
it begins to forward.  This cuts down on propagation of runts,
collisions, etc but can still allow for CRC problems, slightly longer
latency.

The latter two will not isolate errors as well as a switch.  Switch
errors tend to isolate to a port, while hub errors will propagate the
subnet.

And just for grins, remember the 5-4-3 rule - 5 segments, 4 repeaters, 3
populated segments (but that's mostly coax nomenclature).

The "increased" Internet speed may have pushed your network throughput
over the "knee" of a half-duplex performance curve.  You can only get
about 70% of theoretical throughput on a 10Mb half-duplex subnet (if
you're lucky).

But like everyone else said, switch, switch, switch.  Just be a little
careful when selecting a switch.  Also, your cable modem is likely half
duplex as well.

Jeff

Jeff

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *

ATOM RSS1 RSS2