Subject: | |
From: | |
Reply To: | |
Date: | Sun, 13 Oct 2013 22:57:55 -0400 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Depending on the 3000 model, you're only running at 100Mbps, so there is
really no "speed" tweak that is relevant.
You want to insure basic connectivity issues... the 3000 isn't that
great/reliable at autonegotiation, so you may need to hardcode the 3000
*and* the switch on the other side to 100Mbps/full duplex. Nothing
sucks worse than autonegotiation failure... a switch will typically
"default" to 10/half if autonegotiation fails.
Checksums only matter if you care if the data is accurate :) If you
turn them off, errors may go undetected.
If you have enough traffic on the link to really generate congestion,
you may want to check your TCP timers. There have been numerous
postings in the past on tweaking the default timers (which tend to
recover slower than the typical network device on retransmits).
Jeff
On 10/13/2013 3:10 PM, Tracy Johnson wrote:
> The Empire machine is on an admittedly slow network. In other words
> it is on the cheapest Cox business line that was set up for 5Mb upload
> and 1Mb download.
>
> The circuit's real purpose is so visitors in our facility can surf the
> internet over wireless without logging into our network. So the
> Empire machine was put on it as the default endpoint for connections
> coming inbound.
>
> My question it, given the outbound speed is only 1Mb are there any
> arcane tweaks I could change in NMMGR? Would smaller packet sizes
> do? Do I really care about checksum?
>
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|
|
|