Date: 05/12/98 From: Victor Dinh DINH - MIC To: INTERNET GATEWAY INTERNET - SCCB Subject: Single Session LU6.2 Serialization Reference: Note from (EX010898 - SCCB) attached below ------------------------------------------------------------------------ ------- to: cics-l@uga.cc.uga.edu

George, I don't have clues, just some questions and comments.

Is there any entry with trace type KCP SUSPEND in your trace ? Does CICS task for file xfer issue file I/O ? If it does, the next task will get dispatched (or resumed), does some works until it gets suspended,and the next task gets control, ...

BTW, I'm interested in a brief overview of the Tandem code that does APPC to CICS. Would you please provide it to the CICS-L. TIA, victor.dinh@cibc.com My own opinion, not CIBC's -------------------------------- ATTACHED NOTE --------------------------------

Date: May 12, 1998 From: George Sproull [SMTP:George_Sproull@FMSO.NA EX010898 - SCCB To: CICS-L@UGA.CC.UGA.EDU - Subject: Single Session LU6.2 Serialization ------------------------------------------------------------------------ ------ Hi, For our home grown file xfer process, we define 20 single-session LU 6.2 connections using connection/session definition pairs. Each session uses the same modename. The file transfers are initiated on the foreign host (Tandem) and the file transfer process in CICS 2.1 gets a session. A CEMT INQ CONN shows that all twenty connections may be in an acquired status concurrently, and an equal number of iterations of the file transfer transaction show up on a CEMT INQ TASK. Our CICS is always the PLU, but the file transfer transaction is invoked from the Tandem process. Everything seemed to be fine, until we happened to see a trace that showed that on file transfers originating from the IBM system, the first would start (we would see data from the 1st transfer on the trace exclusively), another would start (at this point, no more data would flow on the first session and we would see data from the second exclusively), another would start (no more data from the 1st or 2nd sessions on the trace, only from session #3). When #3 completed, we would see only data from session #2. When it completed, #1 would resume and finally complete. File transfers coming from the Tandem to the IBM have buffers from multiple transfers interspersed (as we would expect). I'm confused, anyone have a clue? (we aren't interested in rewriting for parallel sessions...just curious). Actually performance isn't bad either direction.

Thanks in advance,

George Sproull DISA SSO Mechanicsburg,PA

George_Sproull@fmso.navy.mil