JAS setup?

Our Enterprise server is an AS400 and we run the JAS on an Intel Box.
I've had conflicting information regarding where to store the serialized object tables(F989998/9). Currently we have them on an SQL instance on the JAS machine. I will be re-installing JAS onto two new boxes along with change to IBM HTTP(from IIS) and WAS40 AE (from 3.5 Std).
Do you know if storing the F989998/9 tables on the AS400(ent.svr) as opposed to on SQL on JAS box will have any performance degradation?
If you have a similar setup where do you store your serialized object tables?
If you have two JAS boxes do you have two sets of these tables on both machines ? or do you point to the one instance of SQL? With redirector turned on if one box goes down you can still function if you have two sets of these(F989998/9), one on each JAS box.
If this is the case how do you keep them in sync?
If you store them on the AS400 there is no need to replicate for each JAS box, I think. But, do I give up performance?
Sorry for the lenghty question, I would love to hear your comments.Thx.

Soyer,

We have JAS installes as follows:

The F989999 and F989998 tables are located in the Central Objects for each path code. Multiple JAS servers can talk to a single set of tables.

We use an Oracle DB instead of AS400. Our JAS builds take 1.2Hrs for a full gen. We gen from an Intel box with 2 1Ghz processors and 4Gigs or RAM running W2K with SP2.

I believe that the Redirector is just a method of fail over and load balancing. Which you will get to take full advantage of with WebSphere 4.0. If one machine goes down, the other will pick up the slack. How well it picks it up is dependant on the size of the machines.

I hope this helps.

I am looking at a similar setup. So do you have a seperate maching dedicated to JAS Generation or can you use say your CNC maching. Right now I am generating on the server but I only have one server configured since we are still in the early stage od development. I guess I am trying to find the ideal JAS setup. I was also told by JDE that in order to reduce down time for updating JAS, I should create another datasource for the F98998/9 tables so that you can repoint your jas.in to that source generate the objects restart services and and have jas back up within 5 minutes. that way you will maintain two instances of those tables.

We have two sets of tables. Because of the way that the E-Gen works, you have to gen the tables into a second area. Otherwise you will cut off whoever is logged in. Basically because the E-Gen process truncates the tables prior to rebuilding them.

We store the extra tables in the Control Tables area and then we use an Oracle script that exports the current tables, exports the new tables and then imports the new tables into the Central Objects for public use. We have this process scheduled every morning before people log onto the system.

so you run the script regardless of weather you updated JAS or not. how long does the process take to export and import the tables. Have you ever considered using the JAS.ini to map the tables, or is your process full proof so far

Given our implementation schedule we build packages every day, so we have JAS gens every day as well. The whole process of deploying the new JAS specs takes us an hour and ten minutes. Which we do early in the morning. So far everything has worked perfectly.

How are you changing the JAS.INI? That sounds promissing.

From what JDE told me in the [JDBC URL] section of the INI file you point it to the location if your mapping on the generation machine and generate the objects. then you change the OCM for those 2 tables restart the JAS service and that should be it.

I have yet to test this because I am not doing a lot with JAS right now, but I think it is worth a try if it could reduce down time. Let me know if it works for you if you try it.

That does sounds promissing for the future. However, given that we do builds every day, maintaining the OCMs would become a big hassle for us. I'm going to discuss this with my team and see what they think. I'll let you know what comes out of it.

Thanks!

To Soyer:

We have a very similar configuration. We went through this last year and
this is what I concluded at the time.
I did not try SQL on the JAS server. SQL is a database that you need to
purchase(how much?) or can you get it free?


From 11/5/02 - Posted note.

Based on some answers and questions, I wanted to follow up. Thanks everyone
for your input.
We are running 1 JAS Server: WebSphere 3.5 on Compaq ProLiant 3000 - dual
Intel x86 600 mhz with 2GB Memory/ 30 GB Storage - Win2K SP2. Our
enterprise Server is an AS/400 model 720. JDE XE with Service Pack 20 -
E1. (It meets minimum specs but is certainly not overpowered)
From the feedback and my own experience, this is what I think. When I
generate ALL objects, you end up with about 107,000 records of data in the
F989999 and 220,000 records in F989998. These tables hold the
serialized(JAVA) objects and are on the AS/400. WebSphere access these
tables and then CACHES the info on the INTEL box. Because there is just too
much data, it takes a long time to load the first time(you can see this in
the LOG files.. reading the F989998 file). If I pick and choose the objects
I need generated, it works fine. We have the basics, (Order Inquiry, Stock
Check, Inventory Inq, some reports). Now I only have a few thousand records
of data in the F98... tables and it works fast.
I did apply the Fix OTI-02-0077. This does speed things up but not enough.
We went from a 15 minute wait to a 4 minute wait when signing on the first
time and having ALL objects generated.
The ORACLE fokes out there don't seem to have this problem... perhaps the
all that money spent on the ORACLE database is worth it!!.....
There are several "tuning" options available in the JAS.ini file and in
WEBSPHERE, but I can't seemed to figure them out.
If anyone else is running this same configuration and has generated ALL
objects and is running fast, let us know how you do it.
Thanks

Tom Crawford



Tom Crawford

How about the F98BLOL file? I've tried to do similar things but for ER which are big and don't fit on one record in F989999, it uses F98BLOB for additional.

Wes

I would recommend to use Two sets of F989998/9 and F98BLOB tables and to use WAS 4.0 AE Vertical and Horizontal cloning and have a separate jas.ini for each clone.

Before you perform the eGeneration make sure you are going to generate to the correct table set in [JDBC URL] section in jas.ini. Once you generate just change the jas.ini [JDBC URL] section to pointing to the correct table location for each clone and restart the clone. This case you do not have to import and export tables as well as down time is minimum as you are restarting clones individually.

NOTE: Web redirector or cloning, both single point of failure. To prevent use IBM ND ( IBM Network dispatcher ) comes with IBM Edge Server.