The contention between centralized and decentralized structures is a persistent theme throughout history. This continuous process in search of an optimal structure is dependent on a variety of variables and agendas. Clearly, the process doesn't always result in a humanitarian scenario.
Over the last few decades we have moved from centralized mainframes to more distributed computing models that have pushed intelligence and processing power to client devices.
According to technology analyst Kevin Werbach, over the next decade centralized computing will fade into history. He believes that centralized computing will be unable to scale sufficiently to deal with the complexities and distributed nature of the next-generation Internet and Web services. The most famous examples of that highly decentralized model are the peer-to-peer music sharing services.
Traditionally, centralized computing models have been considered more reliable, secure and manageable. The buck stops in the IT shop and governance is the province of a central command group. Today, many companies have compute resources broadly distributed within the enterprise or outsourced at remote locations, but they are managed virtually by a central group, which itself may be somewhat decentralized.
The issue is finding the appropriate balance between centralized and decentralized computing as well as organizational structures that promote collaboration and nimbleness. But as distributed computing models mature in the coming years, the task of automating, managing and provisioning the network will present a significant barrier to shifting away from centralized controls.
At Werbach's Supernova conference this week in Palo Alto, CA, decentralization provided the backdrop for a discussion on the future of software, communications, and media. I spoke with Werbach about his views on how decentralization and moving more intelligence to the edges will impact the future of enterprise computing.
Tech Update: You have said that decentralization is a fundamental issue for the next decade. If you were talking to the CIO of a large corporation, what would that mean?
Werbach: Companies are recognizing that the business and technology world is moving toward decentralization. They also recognize that they must maintain order and efficiency in a decentralized environment. That doesn't mean everything pushes to the edges -- pressure is going in both directions [centralized and decentralized], and you have to the find the equilibrium point. Companies are increasingly outsourcing and integrating with partners, for example, and they have to move away from traditional centralized command and control structures.
Tech Update: Why are companies moving toward more decentralized architecture and organizational structures?
Werbach: Centralized systems are becoming so complex and are difficult to scale. With commoditized computing and connectivity, the costs are much cheaper than the massive mainframes of past. In addition, as IT becomes more commoditized, the differentiator within organizations is people and finding ways to help them collaborate and get maximum advantage.
Tech Update: It doesn't make any difference to the user population if a mainframe or cluster of PC servers is behind the curtain. Certainly, costs are driving more distributed computing models, but how does the distributed model relate to organizational efficiencies?
Werbach: If a business is trying to communicate to 100,000 employees who march in lockstep, more centralized framework makes sense. But if a business is trying to unlock the potential of employees to communicate with one another and collaborate inside and outside organization, a single, centralized system can't do it.
Every business is finding employees using e-mail and instant messaging as common product not because it's good, but it's the only thing that works across boundaries. You don't have time or money to set up an extranet each time you want to communicate [outside the company]. A great need exists for new kinds of collaborative business software that can provide the same power as more centralized client-server software. But it's not an all or nothing proposition. It is less about the technology and more about the way people will work. Today there is a certain synergy with what business demands and where technology going.
Tech Update: What technologies are driving the trend toward decentralization?
Werbach: The next basic technology for enterprise IT is Web services. Whatever you think about the specifics, this is the direction for enterprise software--a componentized, decentralized model. It's really an evolution in the system architecture of enterprise software to a more distributed model. This evolution will happen over incremental steps, going away from a world where one ERP system is the center of universe to one in which anyone can build and aggregate components to form new kinds of services to meet their needs.
There is a question as to whether Web services will devolve into a war of big vendors claiming to use open standards but going off in their own direction. But it is incumbent upon enterprise customers to let the vendors know that such an approach won't fly. There are also big questions about what the business models are and how they change the way technology is sold.
Tech Update: Does the move toward Web services and decentralized technology mean that centralized models are flawed, or is this just a swing of the pendulum taking advantage of distributed technologies and a more mobile workforce and society?
Werbach: Areas of centralization depend on organization. Some companies will want to keep tight control of what employees do, and in other companies, the last thing they want is exert that kind of control. It's an exercise in figuring out your core competency. Replacing or augmenting centralized systems with more distributed technology will be cheaper and more efficient.
A number of people point out that biology and complexity science are looking at basic laws and patterns that occur in complex adaptive systems, such as ant colonies and the human immune system. That doesn't mean we need to run off and build everything with genetic algorithms, but if we start from the premise of looking at the most complex systems on planet and reverse engineer how they work, we will find many decentralized systems operating in parallel.
Tech Update: How to deal with managing distributed environments?
Werbach: The challenge is for vendors to put in place management mechanisms. There are enough benchmarks for software buyers to evaluate the viability of various solutions.
Tech Update: You say that centralized systems are crumbling under the strain of complexity, but aren't distributed systems inherently complex?
Werbach: There are different kinds of measurements of what complexity means. If you compared the telephone to the Internet when it was first conceived, a lot of engineers said the Internet couldn't work. It has proven to work, and now when people look at the Internet, no one thinks you could have built a network as big as the Internet with a centralized directory. Decentralization is relevant in so many different contexts that the broader implications are often missed.
