My original question:
> I have an application for data acquisition where 8 external devices will be
> sending packetized data over 19.2 KB async modem lines to a single Sparc
> II. The normal throughput will only be about 2400 baud per line, but in
> peak mode all lines may run to full capacity.
>
> The software developer has recommended a Texas Microsystems SSC-80 SBus 8
> port serial mux, which has an advertised OUTPUT aggregate throughput of
> 36,000 CPS, but not input throughput specifications. I think it was
> recommended because it was one of the first SBus async cards to be
> announced. Since most data will be incoming, with only sliding window acks
> and occasional control info being sent to the external devices, I am more
> concerned with input capabilities.
>
> Does anyone have any experience with either this board, Sun's news SBus
> Serial/Parallel card and driver, or other SBus async solutions? Are any
> system better or worse for this input-laden application? I will summarize
> the results.
Responses:
No one had direct experience with either product. I received several
suggested alternate solutions:
------------------------------------------------------------------------
1. From: milt@pe-nelson.com (Milt Ratcliff)
This is not quite what you are asking but is an alternate approach. Livingston
Enterprises(Pleasanton, CA, 415.426.0770) sells a TCP/IP terminal server
which they claim performs 'real' time acquisition quite well(each port to
19.6). They have told me that it is used in a number of hospitals for
real-time monitoring of medical equipment(heart, etc.) feeding a single
Sun. In these applications, the instruments are very 'dumb' and the
server must not loose any data.
We have a pseudo real-time application(or interfaces have local buffering)
but may look at them as well as others. Our application requires that we
use terminal servers since the point of acquisition may be quite remote from
the cpu.
------------------------------------------------------------------------
2. From: dannet!bruce@uunet.UU.NET (Bruce Hill)
Danford Corp. has an 8-port Asynchronous Serial Port SBus card that might
better serve your needs. The product is called the SEU-6900 (Serial Expansion
Unit). I wrote the device driver for this product and our main goal was
to handle 38400 Baud input on all 8 ports simultaneously. We have tested
our product against the Sun and Texas Microsystems boards and blew them both
away. I will send some info to your US Mail address.
You did not state whether your application uses RAW or COOKED input, but
we can handle RAW (-icanon) input on all 8 ports at 38400 Baud.
Please call our Sales Director, Jay Healy at (213) 514-9334 if you have any
questions.
------------------------------------------------------------------------
3. From: Neil Gorsuch INTERNET: neil@cpd.com
The character throughput of our current model of SCSI based
serial/parallel port box is ~30,000 cps output, ~10,000 cps input.
Baud rates continuously adjustable up to beyond 100 kbaud. Uarts have
16 deep fifos on both input and output on each channel. About to be
announced newer model will probably be 3.5 times faster. No new
driver or kernel rebuild needed. Full flow control, including
RTS/CTS/XON/XOFF, etc, etc 8-).
Contact: Uninet, a division of Custom Product Design, Inc
PHONE: +1 714 546 1100 FAX: +1 714 546 3726
----------------------------------------------------------------
Doug Neuhauser Seismographic Station
doug@perry.berkeley.edu ESB 475, UC Berkeley
Phone: 415-642-0931 Berkeley, CA 94720
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:21 CDT