RE: AS400 vs. NT Enterprise Server - Why Not Both?

[don_schoen]

RE: AS400 vs. NT Enterprise Server - Why Not Both?

List

We are an AS/400 E/S shop about to grow to 120 One World users using sales
order, financials, and warehouse management. We have almost everyone
connected via Citrix. The 400 is a 2-way 720 and over 50 spindles.

We believe that the workload on the 400 may be much easier to manage if we
move custom financial reports to a report server, and possibly move order
import processing (which includes advanced pricing) to an application
server. In terms of first shift, that would mean that big batch jobs that
could impact interactive users would be on a different platform, and would
only contend for data base services. The rest of the workload would be more
homogenous, and could be managed in a few subsystems.

Has anyone any real experience with this approach?

We end up with 3 questions:

1. could the batch jobs we were worried about flood the 400 with IO
requests that get a higher priority than we would have given that batch
processing had it taken place on the 400? That is, did we lose control over
that batch job by putting it on a app server or fat client?

2. is there any way to adjust the priority given to the data base
services to the fat client or the application server - so that the flooding
wouldn't impact us negatively?

3. is the introduction of an application server (in this case an N/T
machine) generally a good way to off-load 400 batch processing load, and
does it really help overall through-put? And by how much?

Thanks for any input.

Don Schoen
Terlato Wine Group
Xe Update 3, SP14.2

 

[Tom_Davidson]

RE: AS400 vs. NT Enterprise Server - Why Not Both?

Don,

We currently use an AS/400 enterprise server with about 850 Xe users (100
concurrent). Our experiences below


The only time the 400 gets contention is when I do package builds and the
PAKTOTAM jobs cause an inordinate amount of paging.

server or fat >>client?

I don't understand this question. You simply have to control it a
'different' way. Our experience is that jobs that are 'quick' and don't
move lots of data perform OK on the app server, mostly due to
communications.

flooding

Fat clients and App servers attach to the AS/400 via ODBC (Client Access).
These connections are handled by the QZDASOINIT jobs, their priority is
normally controled by class QSYS/QPWFSERVER.


I can't truly answer this as we have used the app server mostly for
'playing'. The administration of multiple app servers is our main issue,
mostly in teaching our users how to 'find' their submitted jobs.

Hope this helps.

Tom Davidson
Xe Update 1, SP 16.1, ES: AS/400 V4R5, CO: SQL 7.0, TSE with Metaframe
1.8. 2 instances: English/Western European & Japanese.



OW 7332 SP 11.3VER, NT 4.0 SP 5, TSE 4.0 SP 4, Metrframe 1.8, CO SQL 7.0

 

[Jack_Crouch]

Good questions. I stopped the implementation of an NT app server after I realized these things. I am concerned about the priority issue (as well as jobq routing, maintenance, and visibility) with an app server.

Since the QZDASOINIT jobs are prestart... I do not know how you would assign priorities per "future" app server/user.

I have had the experience that running things locally runs quicker than submitting the job to the AS/400. I generally chaulk this up to the priority issue. The difference has generally been noticeable but negligible.

Lately... I have had some really bizarre instances of UBEs taking a nose dive on performance (and the local ones will run fine)... I do not know what is going on yet. Some kind of (recent) problem to solve here...

AS400 V4R5, XE+XU1+35ESUs, SP16, NT-SQL7 for CO