Marian
I'm somewhat rusty on the OSA-Express but I see from the OSA-Express Modes
of Operation web page
http://www.redbooks.ibm.com/abstracts/tips0088.html
that the OSA-Express 1000BASE-T defined as OSE supports SNA. My experience
with OSA-Express has been exclusively with the GbE feature which supports
only OSD and TCP/IP. Thus you are using 3172 emulation and VTAM XCA
definitions with OSE.
I'm afraid I am also not familiar with the Cisco machine so I don't know
what that emulates. You need to compare whatever the Cisco machine emulates
with the OSE mode of operation of the OSA. As you said, the same VTAM
switched definitions and the mode table entry are used for both so the
difference cannot lie there.
I expect that you need to improve the performance parameters associated with
the 802.2 connection over Ethernet.
The 802.2 parameters are as follows:
N1 - maximum size of I-frame
N2 - maximum number of retransmissions following T1 timeout
N3 - maximum number of frames received before acknowledgement
Nw - number of acknowledgements needed to increment Ww when less than TW
RW - limit to maximum number of unacknowledged frames that can be sent
T1 - maximum time allowed for an expected acknowledgement
T2 - maximum time to delay before sending an acknowledgement
TI - maximum time to allow no traffic to flow
TW - maximum number of received unacknowledged frames that can be sent
The performance parameters are N1, N3, RW, T2 and TW. Nw also affects
performance when frames are being lost. (Ww is the "working window" which is
set lower dynamically when frames are being lost.)
Performance over the wire or though the aether is always about minimising
the effect of protocol headers - a point perhaps Lizette had in mind - and
minimizing the effect of flow control - pacing, acknowledgements and so on.
If you had NCP and token ring rather than Ethernet, you could actually
specify the necessary parameters through your switched PU statement
definition. I don't believe all the parameters apply to an OSA defined with
XCA. Thus you need to configure some of the parameters using OSA/SF - if my
memory serves me well.
You can influence N1, RW and TW through VTAM or VTAM application
definitions. N3 and T2 must be configured in OSA definitions.
The size of frames to be sent, N1, is probably going to be determined by
JES2. I see you have not specified a value for RU sizes in the mode table
entry RUSIZES operand. The RU size determined by JES2 will translate, after
protocol headers have become attached, to the frame size, N1.
Note the MAXDATA operand is ignored since your connection to the AS/400 will
use type 3 XIDs.
Thus we can concentrate on the acknowledgement flow, RW and TW. This is
controlled by MAXOUT and 7 is probably far too low. You should try to ensure
that 127, the maximum, is used instead. This involves setting MAXOUT=127 and
also you should be sure that the AS/400 adapter is defined to use 127 also
since TW is the minimum of the RW received in the type 3 XID and MAXOUT.
When I used to have test systems to play with, the 16M 3 or 4 segment token
ring I used needed MAXOUT=20 to be sure acknowledgements were never a
limitation.
It may be that the difference between Cisco and OSA is that the Cisco
machine managed to ignore your far-too-small MAXOUT somehow.
N3 and T2 are concerned with when an acknowledgement is sent. In principle,
for raw speed, you should set N3 to 1 and T2 to 0 - both these values mean
the same thing really, namely that each received frame is acknowledged
immediately. However you will put less traffic on the LAN if you delay the
acknowledgement somewhat giving a chance that the acknowledgement can in
effect be carried in a frame in the opposite direction if one appears in the
interval defined by N3 and T2.
I see you also have far-too-small pacing specified in your mode table entry.
You should specify 63 for all your pacing fields. There are rules over which
pacing fields/operands are used but they are far too complicated, depending
as they do somewhat on the configuration. You can't go wrong if you just
specify the same value everywhere a pacing value is needed and 63 is the
maximum. This assumes you need pacing and with NJE you do. If the traffic
were self-paced with definite responses as 3270 traffic usually is (but not
an LU type 1 printer), you need to specify 0 everywhere a pacing value is
needed.
A few more comments on your definitions while I'm here:
- IDBLK and IDNUM are redundant when you use CPNAME.
- DATMODE is relevant only over an SDLC connection.
- PASSLIM is relevant only for an SDLC multipoint connection (and token ring
when being shared by SNA and IP traffic).
- XID=YES is superfluous in a switched definition.
- SSCPFM is irrelevant for the definition of an SSCP-independent LU, that
is, an LU defined with LOCADDR=0 which, in any case should be defined using
a CDRSC statement rather than LU LOCADDR=0.
- Did you know that you could be using the CP LU as your single
SSCP-independent LU? Of course DYNLU=YES implies you have multiple
SSCP-independent LUs and you have chosen only to define one explicitly. Only
you know what your definitions mean.
Chris Mason
----- Original Message -----
From: "Slabý Marián" <[EMAIL PROTECTED]>
Newsgroups: bit.listserv.ibm-main
To: <[email protected]>
Sent: Wednesday, March 14, 2007 2:16 PM
Subject: Re: NJE poor speed between AS/400 and z890
We use 1000Base-T card and it runs in 1Gbps mode.
There is no big traffic on this Ethernet segment.
Our VTAM definitions:
SN11AS4 VBUILD TYPE=SWNET
PAS400 PU ADDR=01,
IDBLK=056,
IDNUM=28F11,
CPNAME=NESOBA80,
NETID=ITBIC000,
XNETALS=YES,
ANS=CONT,
DATMODE=FULL,
DYNLU=YES,
MAXDATA=265,
MAXOUT=7,
PACING=7,
PASSLIM=7,
SPAN=LANMN,
XID=YES,
VPACING=7
UAS40000 LU LOCADDR=0,
SSCPFM=USSSCS,
MODETAB=MODTAPPI,
DLOGMOD=APPCMODE,
ISTATUS=ACTIVE
UAS40001 LU LOCADDR=01,
LOGAPPL=JES11,
MODETAB=MODTNJE,
DLOGMOD=AS400NJE,
ISTATUS=ACTIVE
LU UAS40001 is for NJE.
DLOGMOD Definition:
AS400NJE MODEENT LOGMODE=AS400NJE,
TYPE=1,
FMPROF=X'03',
TSPROF=X'03',
PRIPROT=X'72',
SECPROT=X'72',
COMPROT=X'4020',
SSNDPAC=X'00',
SRCVPAC=X'03',
RUSIZES=X'0000',
PSNDPAC=X'03',
PSERVIC=X'00',
ENCR=0
Only note - these definitions work fine by CISCO router.
Marian Slaby
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
