brown@noao.edu (Mike Brown) (01/20/89)
Problem:
The effective throughput of a 56 Kbps line appears to be 19 Kbps
when measured using a "print" test. Does anyone have test results
to prove that an IBM RT (model 125) can really drive a 56 Kbps SNA
synchronous circuit at 56 Kbps using any of LU2, LU3 or LU6.2?
Configuration:
AIX 2.2, SNA services, Rabbit SNA 3270 Plus,
Multiprotocol Adapter, IBM RT Model 125,
56 Kbps Digital Leased line to IBM 3745 VTAM controller.
Unsupported configuration:
V.35 56 Kbps DSU/CSU is connected the RS232 port #1 of the
multiprotocol adapter. (I've recently found out that
X.21 is the only supported way to run at > 19Kbps....)
Description of Tests:
A locally written MVS utility sends a file to printer logical
unit. The time required to print the file to an emulated 3287 mod 1
printer on the RT is measured using Netview on the MVS host.
The Rabbit SNA 3270 Plus software allows the RT to emulate a
3274 controller that runs the 3287 printers.
The file size in 8 bit bytes divided by the transfer time yields
the effective throughput. The timing uncertainty of the
session start/end of Netview is insignificant.
I immediately suspected that the "printer protocol" was the
performance bottleneck. Using netview on MVS we watched
two frames sent to the RT and then a pause until the RT
acknowledged the "page". We tried printing to TWO different
printer logical units on the RT in the hope that the effective
throughput would double. Unfortunately the time required to
simultaneously print a file to two printers doubled. I.E.,
the effective throughput did not increase by using two
printers at the same time.
Summary:
I have run loop back tests on the RT with one and two multiprotocol
adapters connected together at 19.2Kbps (externally clocked) with
no significant performance degradation using LU 6.2. I thought
that I had verified that I had at least an aggregate throughput
of 36 Kbps.
I don't know if the multiprotocol adapter hardware is the bottleneck,
the VRM device driver(s), AIX SNA system or the Rabbit SNA 3270+ code.
Or perhaps the mess all together...
I went to a great deal of effort to have a 56 Kbps circuit installed
to allow throughput testing of a "host" computer using LU 6.2 after
being assured by internal ibm technical people and ibm sales people
that the RT multiprotocol adapter could drive a 56 Kbps line at 56 Kbps.
IBM software support is useless because of their attitude that if I
don't run the system exactly as stated in the documentation then
I'm on my own. I'm fed up with the attitude of
"nyah nyah we don't support that..." and have been told that this
will be the response since I am using the RS232 port on the
multiprotocol adapter instead of the X.21 port.
Never mind that the restriction of which port can run at
>19.2Kbps is buried in the VRM device driver technical
reference manual and no where else....
Do I sound frustrated? You bet! Any advice or constructive comments
are welcome. I'd like to get the performance I thought I had bought.
Regards,
Mike Brown Business Computing Group
Information Systems
Southwestern Bell Telephone Co.
One Bell Center, Rm. 24-V-5
St. Louis, MO 63101
(314) 235-7863
uucp: ....!sw1e!umsgmab
internet: brown@wucs1.wustl.edu or brown@noao.edu