Introduction

Lately, I've been pondering something. I've come across projects like "legacy migration" and "host migration" for quite some time, but I feel like many of them are a bit too showy or gimmicky.

There could be several reasons for this, but this time around, I want to explore what kind of migration would be truly ideal, incorporating some wishful thinking and imaginative ideas.

What is Legacy Migration?

Legacy migration involves moving a dated computer system to a modern and clean system.

The majority of legacy systems were created in the early days of computers, like "general-purpose machines" and "mainframes". Nowadays, only a handful of companies, such as IBM and Fujitsu, provide these systems. Recently, there have been announcements that Fujitsu plans to withdraw from these systems in 2022 and 2030.

The METI DX report describes legacy systems as aging, bloated, complex, and black box systems, and this is the key reason for considering migration.

What is Migration?

As mentioned earlier, the goal of legacy migration is to address the lack of support due to vendor withdrawal or other reasons, and to get rid of the complications, complexity, and hidden issues that have built up from numerous modifications since the early days of computer systems.

However, many of the migration projects I've come across often involve:

1. Shifting from an outdated general-purpose machine (no longer supported) to a different general-purpose machine (currently supported).
2. Automatically converting COBOL programs from legacy systems to new systems using tools.

Will this really address the core problem of legacy systems, which is the reason for undertaking the migration project?

Regarding (1), there's a global trend suggesting that legacy systems will gradually fade away over several decades. While they haven't vanished completely, their prominence is decreasing. It's unlikely that general-purpose machines will regain their position as the main platform in the future. I believe (1) is just a temporary fix until the next round of support expires.

As for (2), issues like bloat, complexity, and black box won't be resolved at all. In fact, using tools to automatically convert data might even intensify the problem of black-boxing.

In my view, this approach contradicts the initial purpose of legacy migration.

What is an Ideal Migration?

In simpler terms, it's a move that tackles the main issue of the legacy system—the very reason behind initiating the migration project.

However, in practical terms, we think that merely carrying out a comprehensive system migration may not completely fix the problem. We believe it's essential to complement it with a business-level migration, incorporating the use of that system.

Why can't systemic migration alone solve the problem?

For an ideally systematic migration, the following are essential for it to be succesful:

1. Understand the complicated and outdated legacy system that's the source of the migration.
2. Improve and construct the modern system that will be the migration destination.

Those engineers who can tackle task (1) are getting older or retiring, and it's not very appealing for younger individuals to willingly learn this field. So, finding people to handle it is challenging. Additionally, it's even harder to find engineers who can build a modern system for task (2).