Softroad
  • contact
  • menu

migration-pattern

COBOL to Java

Legacy Is A Big Obstacle for Advancement

Enterprises/organizations who has HOST legacy systems are trapped in the legacy technology. It makes it hard for them to take advantage of the latest development in IT technology/trends/solutions, such as cloud computing, digital transformation, etc. This becomes an obstacle for both companies and individuals to keep their competitive edge over other players.

COBOL to Java Conversion Methods

  1. Straight conversion: Mechanically convert existing systems as-is.
    　　　　　　　　　　※The migration method is used for a long time.
  2. Generative AI conversion: LLM makes effective conversion and technicians
    　　　　　　　　　　　　　fix bugs and test.
    　　　　　　　　　　　　　Generative AI conversion
  3. AI-powered migration tool: Use AI tools not LLM to understand COBOL
    　　　　　　　　　　　　　　 programsand generate correct pure Java code.
    　　　　　　　　　　　　　　AI-powered migration tool

Problem with Straight Conversion – Conversion

Straight conversion or line-by-line conversion is not good for legacy modernization. Straight conversion or line-by-line conversion is prone to creating bugs. Straight conversion or line-by-line conversion creates big Jave foundation with many special Java functions. Straight conversion or line-by-line conversion creates new system processes like legacy system. Straight conversion or line-by-line conversion leads to performance issues. Customers have to deal with these issues manually by themselves.

Problem with Straight Conversion – Testing

Straight conversion or line-by-line conversion vendors tend to test only 1/3 or less of the migrated code rather than 100% coverage on testing of migrated code. Straight conversion or line-by-line conversion vendors only test as much as the data provided by customer covers. Comprehensive test is left with customer or 3rd party vendors. Customers have to deal with hard to fix performance problems.

LLM (Popular general-purpose generative AI model) and Its Problems

LLM Application

　Use generative AI tools like ChatGPT to convert COBOL to pure Java systems, C/S systems to WEB, etc.

LLM Limitation

　1. It requires learning. No direct injection of knowledge.
　　・Extremely large amount of high-quality training data v.s. insufficient amount of COBOL and
　　　Java available for free.
　　・LLM evaluation requires extremely fast processing v.s. CPUs are too slow to meet the demand.
　　　Difficult to evaluate by COBOL/Java comparison tests.
　2. Additional learning for each round of bug fixes changes conversion rules.
　　System test must be done all over again.

Low quality conversion that requires huge manual workload of debugging, adjustment and retesting.

System Reform Conversion： Language Conversion Using Knowledge-injected AI

SoftRoad has big Knowledge DB (it is like a big dictionary) which is built upon our 20+ years of experience. Our tool use the legacy code (COBOL, RPG, etc.) dictionary to read source code of legacy system and understand the underlying logic. Then, our tool uses Java dictionary to create native Java code as if it was written by human programmer. In this process, we can change data structure to RDB and CSV file, we can even convert to customer’s own framework. We use open style SQL to process data, to solve performance problems. We can expand ID digits, etc. We are the only company using AI powered tool to do 100% automatic conversion with high quality & maintainability.

QA of System Reform

Complete verification of all technical elements

100% coverage of Technical Points. Test is automated as well. We use our test tools to input data, run the programs step by step, compare the output, and find the places where the discrepancies happen. We are the only company who can identify every detail grammars as technical points and find out where and how many times a particular grammar is tested. That is how we achieve 100% test coverage, instead of leaving that responsibility to customers.

Test with actual business data

We run the migrated system using data from real business day. Then we compare the output, until they are the same. This serves as a cross check to make sure no particular data or program problems are left untreated. This gives customer peace of mind.

GUI/CUI migration (WEB)

GUI/CUI migration to WEB system

High Value-Added with System Reform Technology (cases)

Case study: Adding value while migrating legacy system

Success Stories

Client Overview

    This case was reported by Nikkei XTECH earlier this year, with title “ITOCHU Foods aims to break away from Fujitsu general-purpose machines by converting 3 million steps of COBOL to Java”.
    https://xtech.nikkei.com/atcl/nxt/column/18/00001/08840/

ITOCHU Foods
One of the largest alcoholic beverages and food wholesaler in Japan
Annual Revenue for FY 2022 – over USD$4B

Itochu Foods searched for an IT vendor with a proven track record of handling many legacy migration projects and decided to work with SoftRoad.

Project Overview

Goal
・Eliminate Fujitsu’s general-purpose machines on which Accounting, Sales and Logistics systems/applications operate
・upgrade COBOL applications running on general-purpose machines to Java

Legacy system
The number of COBOL programs for accounting systems: 7,000 to 8,000
Line of source codes: more then 3 million.

Target system
The new accounting system uses
・Amazon EC2, a virtual machine environment from AWS,
・SQL database service “Amazon RDS” and object storage service “Amazon S3” as databases, and
・to execute batches using “JP1“.

The migration of accounting system completed in August 2023 was the first step.
The remaining systems will be updated in spring of 2026.

Result

・”Almost no performance problems occurred” says Headquarters Director Namimoto.

・SoftRoad President Lewis explained, “Java is often said to be slow. However, if you convert it properly using AI and tuning special places by engineers, you will be able to achieve processing speed that Java should have had.”

・SoftRoad helped save 50% of migration cost for customer.

・Plus, ITOCHU Foods does not need mainframe anymore. They will save considerable amount of cost of running mainframe each year.

System Reform Schedule

Migration/modernization project process and timelines. Steps included in the process are: general quotation, inventory/slimming, define required specification, detailedd quotation, migration design,summary of existingfunctionalities, development, migration test, functional design, development, integration test, system test.

Migration Cases

Six systems of two IBM mainframes such as production, sales, and after-sales service in a certain manufacturing industry

Development process
Inventory, migration design, source code migration – system test, user acceptance test support, go-live
System reform information
COBOL to Java
COBOL to MF COBOL
NATURAL to Java
PL/I to MF PL/I
JCL to Java+XML file
ASM to MF COBOL
ADABAS to SQL Server
OS/390 V2 to Windows Server 2012
Current system scale
COBOL：2,487KL, JCL：545KL, NATURAL：429KL, PL/I：2,156KL, ASM：35KL, Form：781, Screen：1,572
Development period
20 months(* Migration of each system was carried out in about one year.)

Manufacturing System

Development process
Inventory, migration design, specification documentation, UI design – migration test, system test & user acceptance test support
System reform information
COBOL to Java
PL/I to Java
JCL to Shell
VSAM to Oracle
host to Linux
Current system scale
COBOL：141.5KL, PL/I：37.2KL, JCL：44.8KL
Development period
13 months

New Core Accounting System

Development process
Inventory, migration design, source code migration – migration test, functional modification, system test, user acceptance test support, go-live
System reform information
MED, PSAM to HTML5.0
COBOL to Java11
ASM, Utility to Java11
JCL, Macro, CLIST to Java11 & XML file
File(PS, DAM, PO) to Linux standard file
VSAM file to SQL Server 2019
XSP to Red Hat Enterprise Linux 8.4(AWS)
Current system scale
MED, MAP：11.9KL, COBOL：1,934.2KL, JCL：751.3KL, ASM：3.8KL, CLIST：1.7KL, COPY：41.7KL, Utility：50, Macro：4KL
Development period
30 months

Computer Aggregate Processing System

Development process
Inventory, migration design, source code migration – migration test, system test & user acceptance test support, go-live support
System reform information
IDLⅡ, COBOL, COBOLS, HPL to Java
JCL to Java & XML file
ADBS/RIQS to SQL Server 2016
ACOS4 to Windows Server 2016
Current system scale
IDLⅡ：775.9KL, COBOL：386.8KL, COBOLS：49KL, HPL：9.8KL, JCL：327.3KL
Development period
36 months

Next Generation Core System

Development process
Inventory, migration design, source code migration – migration test, system test & user acceptance test support
System reform information
YPS COBOL to Java
ASM to Java
NDB & RDB & VSAM to Oracle 12c
Fujitsu host to RHEL6
Current system scale
COBOL：520KL, ASM：0.9KL, JCL：178.5KL
Development period
11 months

Cost Management System

Development process
Inventory, migration design, specification documentation, source code migration – migration test, system test & user acceptance test support
System reform information
COBOL, COBOLS, COBOLX, IDLⅡ, RPG, Macro to Java
JCL to KShell
ADBS to Oracle
VSAS/RIQS to Oracle
ACOS4 to Linux
Current system scale
COBOL：104.3KL, COBOLS：124.5KL, COBOLX：7.7KL、DATASSF：5.0KL, IDLⅡ：391.1KL, RPG：1.4KL, JCL：431.5KL, Macro：10KL
Development period
20 months (* POC included)

CONTACT

For more details about system migration,
please click here.