E9.1 AS400 Data Migration to cloud

Skhandelwal

Member
Hi All

We are planning to move our DB2 database(on premise) to somewhere in cloud(ideally a Azure SQL server). Rest of all our JDE servers are windows and can be easliy moved to the cloud via VMotion.
I have few questions and would really appreciate any insights:
1. Is there any way we can lift and shift our current AS400 DB2 database to a cloud solution..?
2. What is the best way to convert JDE DB2 data to MS SQL..? Use a third party software or use JDE's regular conversion methods(R98403)..?
3. If we use a third party software to convert the DB2 data to SQL, what needs to be done within JDE..? Just create a plan and add a new DB server and then build a full package..?

Here's our configuration:
JDE Application Release 9.1, Tools Release 9.1.5.4
DB Server: AS400 7.1 i series
JDE App server, JAS server etc all WIndows

Regards
Saurabh Khandelwal
 
In general I recommend R98403 / R98403A are used and run in parallel when doing platform conversions since it is the only way to say you did it in a "supported" way. With large tables I have used various tricks with datasources and database views in the source to cut the F0911, Sales Ledger, Cardex and other large tables into chunks that can then be processed in parallel by multiple versions of R98403A. R98403A can also be cloned and hacked to run on servers and with JDE on Wintel being able to use ODBC you can have the new SQL server pull data from the AS400.

That being said sometimes there is just too much data to cut over in a reasonable go-live window. If you use R98403 to prepare your target database with the "correct" table structure for the MS SQL platform you could then look at using other ETL tools to transfer data.

In either approach you need:

1) A thorough data validation approach from both a database level (record counts, NULL checks, etc) and a functional level (trial balances, customer statements, inventory valuation). The artifacts produced by validation serve as audit proof that your new system is s true clone of the original.

2) Lots of dry run testing to establish the timings and validate a correct final result
 
To add to the great points by Justin, depending on the size of your data you also want to consider the time and bandwidth required to copy this data from your on-premise to Azure. Do you have some sort of dedicated connectivity to Azure using something like Express Route?

A common approach for these types of platform conversions combined with cloud migrations would be to first convert the data from AS400 DB2 to SQL Server on-prem (By on-prem I just mean where ever the current servers are) and then copy the compressed SQL backup file to Azure and restore there. Identify the pieces of data that you must keep refreshing to Azure (Typically Business Data and Control Tables) and others that you might possibly freeze after a few initial copies and keep in sync through dual maintenance or other methods (security/objects (via PAR files etc).

The end goal would be to minimize the activity for the final go-live cut over and you will need to do a few mock runs to get all the timings down to a tee.
 
I would run R98403 if the database size is not large. R98403 might not make the cutover window if the database size is large.
I would look at data replication tools like golden gate for a large shop. But if you find the license fee too high, then take the approach mentioned by ice_cube to break down the data set.
 
Thank You everyone for the response.
Any insight on what needs to be done for adding a new SQL server in JDE..? Does anyone have any step by step document for adding a new data server in JDE..?
 
Have a look here:


Thanks Justin. So I did try to follow this document and created a Add on Server plan(just for data server). I am lost as to what needs to be done next..? Just go to the database data sources screen and create new data sources pointing to the new SQL server..? And do this for planner, system and server map..?
Also, do I need to run the PPack for the data server..? We already created the required JDE schemas and converted the data through SSMA.
Any help is highly appreciated.
 
option 1 - certainly not. Lift&Shift strategy is far from efficient solution in such cases.
 
Back
Top