Is it reasonable to use Obsolete Platforms as part of SOA?

-->

One of SOA's Value Propositions is platform independence. Some people's view is that SOA is a collection of Lego like services connected together as Black Boxes so it does not matter, which platforms and technologies are used for service implementations. Although the Lego analogy could be useful, it is sometimes too simplistic.Usgae of Obsolete Platforms as part of your SOA can demonstrate the limitations of platform independence in SOA implementation.

Obsolete Platforms are platforms near the end of their Life Cycle. They are characterized by lack of new installations, gradual migration to other platforms and slower development and updating of the platform's software.

In previous post titled Mainframe and the Dinosaurous Myth, I discussed Mainframes as a full participant in the SOA implementation. Make no mistakes, Obsolete Platforms are not z/OS, CICS or DB2 and therefore probably will be at best partial participants.

Technical Considerations

The main technological considerations are:

• The applications are Monolithic Legacy applications.Identifying services within these large monolithic entities is a very difficult task. Extracting services is usually even more difficult, requiring complex reengineering or wrapping entities. The wrapped entities may be too large including more than one service.

• The architecture is Tightly Coupled so implementation of Loose Coupled SOA is costly and difficult. Loose Coupling is not a must, however the majority of modern SOA implementations (especially Web Services Framework SOA implementations) are Loosely Coupled.

• Lack of SOA products and tools

Most vendors' products support only frequently used platforms. Supporting Obsolete platforms

requires technical efforts and building marketing and sales channels. Potential revenues may be too small

for justify these extra expanses.

• Limited support of Web Services

There are three elements of "Limited support" in this context:

1. Fewer Web Services standards are supported

2. Current Web Services standards version may not be available

3. Implementation is not based on code generation and may require complex and potentially error prone

coding by infrastructure programmers or application programmers.

Key Considerations

These considerations are usually more important than the technical considerations.I will demonstrate it by example.

Many years ago my responsibilities included Capacity Planning and Hardware upgrades decision of an MVS (the origin of z/OS operating system in the eightieth and ninetieth of the previous century) Mainframe installation servicing more than 3,000 users, therefore I knew quiet well Mainframe computers capabilities and prices.

As a consultant to a Data General (DG) installation using its proprietary operating system, I was surprised that the price of a new computer was a lot higher than the price my installation paid for IBM Mainframe. The number of users in the Data General installation was a lot smaller than the number in the MVS installation same as the number of online transactions so the required computing power (Processor, IO, and Memory) in the Data general installation was a lot less than the power required in the MVS installation.

Why was the DG computer so expensive? It was an Obsolete Technology.

The price tag was based on two factors: It was a n Obsolete technology so the company did not plan to sale other computers in two or three years after that deal and the customer was locked in, so he has no other choice.

Unsurprisingly, the price of more powerful DG UNIX server was about 20% of the price of the proprietary DG server. The customer had a choice (IBM, HP, Sun etc.) and the vendor expected to sell additional computers.

As illustrated above, the first consideration is high Software and Hardware costs. The second is high maintenance costs (including maintenance fees and electricity). You also should expect monotonic growth of maintenance fees.

The third consideration is diminished skills.

The bottom line is that a SOA implementation based on Obsolete Tchnology will be a costly and a slagish implementation

What is the real Value Proposition, if any?

The real issue of Obsolete Technology based systems is a choice between bad and worse alternatives. You should choose the least harmful alternative.

In some cases in spite of the cited above technical limitations, it could be SOA implementation based partially on Obsolete Technologies.

The advantage of it in comparison to Big Bang migrations or continuous deployment of Monolithic Obsolete systems is the ability to migrate gradually and smoothly by taking advantage of the relatively independence on platforms and Service and Consumer loose coupling.