Serial ISDN to VMS system

• Previous message: nikki_wire: "Re: a little scripting challenge"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Date: Mon, 23 Feb 2004 17:27:12 +0000

As I've had success with this now, just posting information of how it
was done for posterity.

Summary: In the back-of-beyond, I'm using mid-band access, that is 64k
and 128k ISDN to my ISP. PC access is fine, the problems arose when
trying to get good data rates to the VMS systems, the software
environment is irrelevant.

I'm using a Courier-I (ISDN Terminal Adapter / Modem combined) and a
DS700-16 which is partial modem control.

I've wired a DB25, shorting the RX and TX grounds, passing modem signals
for hardware flow (default modem port setting). The Courier-I is in V120
mode, and 115200 kbs. Terminal server has a remote LAT created port for
dial-out. Previously I was seeing low data rates, aggregate of 2.5kbs
when I was expecting at least 6kbs. The modem was showing high flow
control operations (RTS/CTS signalling very active). I'm using dynamic
PPP from the host to a LAP port (not IP on the terminal server).

Anyway, I fell over a description of the Circuit Timer parameter on the
DECserver, and I found myself thinking that PPP was based on small
packets, probably less than a buffer's worth, so I reduced the default
value of 80ms to 20ms, and I now started to see data being "streamed" at
64k with little interaction from the hardware flow control.

So that was the answer. I'm going to try a lower value still, but in any
case, this is a vast improvement.

In case you're interested, the circuit timer is a time, in milliseconds
that a transmit to the host is performed, even if the actual packet is
not up to the "transmit" size. Effectively, I believe that a PPP packet
failed to hit the transmit size, so it hung around in the DECserver,
until the circuit timer expired, and the packet was sent (padded with
nulls), ready for the next transmit. Now I know this, I'm probably going
to see what else I can experiment with in that area (i.e. timers on the
host).

Of course this information may be useful to anyone else who may be
struggling with low data throughput with small asynchronous data
packets, but you may also already know this.

-- 
Regards, Nic Clews a.k.a. Mr. CP Charges, CSC Computer Sciences
nclews at csc dot com

• Previous message: nikki_wire: "Re: a little scripting challenge"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]