The software industry has always seen that part of its role is to recommend (sometimes forcefully) the use of new architectures and design principles to take advantage of recent technology innovations. However, it hasn’t always taken the lead in adopting these practices internally in the way it builds its own products.
The reason for this is that, at the lower layers of the infrastructure stack, software becomes increasingly performance-critical. Architectures that are designed to improve the quality and/or the speed of deployment of business solutions usually have some sort of performance trade-off that is acceptable at the highest business-visible layers, but becomes unworkable if implemented in the underlying software platform.
This is certainly true of service-oriented architecture (SOA) – it is very strong on delivering business value in a remarkably short timescale, but does so at the expense of introducing additional layers of virtualization software that add a certain performance penalty.
BEA has recently been talking up its strategy for delivering new and enhanced infrastructure products to meet customer expectations of timeliness and versatility. The architecture for executing on the strategy is called microservice architecture (mSA), and is an adaptation of the principles of SOA to meet the challenging requirements of performance-critical infrastructure products.
The aim behind BEA’s mSA is to give the company a competitive advantage in being able to deliver software infrastructure products more rapidly, addressing changing requirements, and even being able to create products for niche markets that would not be financially viable in a traditional development model.
It uses a backplane that provides registry and repository functionality, but at the microservices level. This provides the lifecycle management controls needed to support the full independence of microservices. The architecture does not, of course, implement the concept through resource-hungry web services or XML messaging, but rather uses a proprietary internal integration technology. It does permit microservices to be loosely coupled so that they can be re-combined in many different ways.
Historically, infrastructure products have always been built through the integration of discrete services, and have always increased in scope over time. For example, a relational database product could be viewed as the combination of I/O management, caching, SQL parsing, thread management, logging, optimization, etc. The enhancement of these individual components, and the addition of others (such as security and XML capabilities in the case of the relational database), lead to the need for considerable integration testing with consequently prolonged product lifecycles.
BEA’s mSA provides a means to decouple the dependencies, providing the means to shorten product delivery cycles and to become more responsive to enhancement requests. It also gives the potential to remove functionality that is not needed in a particular deployment scenario, reducing the product footprint and targeting different market sectors.
BEA has already started the process of breaking down the functionality of its products into microservices, and plans to complete this process by the end of 2008, when it will build all new products in this way. This approach also opens the way for simpler embedding of third-party technology provided by partners or through open source licensing.
This is a far-reaching and challenging strategy for BEA to adopt. It will need to demonstrate that it can create composite infrastructure products that perform as well as those built in a traditional manner. Equally significantly, it will need to demonstrate that it can manage the complex lifecycle issues to deliver stable and functionally-relevant products within an aggressive timescale. SOA has already shown that internal company politics are as challenging as the introduction of the technology, and BEA will need to address the technology ownership and partnership relationship issues, as well as potentially needing to restructure the sales force to compete in new market spaces.
However, if BEA can meet these challenges it will establish a set of capabilities that its competitors will be forced to match. In the process, it will have created a leadership potential of a year to 18 months ahead of its rivals.
Source: OpinionWire by Butler Group (www.butlergroup.com)
