Requires transaction and synchronization support. Possibility to loose of integrity.

 

Data centered arichitecture is a single point failure architecture

 

Scalability

 

Security – which client has access to what data

 

Database cab be the performance bottleneck

 

Yes, but at the same time distributed systems can be very unreliable as well. Performance bottlenecks can gather in areas of the system that are more frequently used than others. Denial of Service attacks plague even the most sophisticated Internet sites, even though their back-ends are made up of many distributed servers.

Many say that the days of the IBM mainframe are gone, but cost of distributed systems are driving customers back to a more centralized system. Ultimately, there really isn't a silver bullet, end-all solution. Every use case requires its own specific analysis, which leads to a particular implementation that might involve a combination of both a centralized and de-centralized architectures.

I firmly believe that we shouldn't firmly believe. =) What I mean by this is that every situation must be viewed from a perspective that isn't hampered by our own personal beliefs. There are some instances where a centralized architecture may be more beneficial and other circumstances where a decentralized architecture would be more desirable. It can be boiled down to essentially two words: No Dogma.

~ Emory

can not agree more. Each style has its advantages and disadvantages. Usually the choice is not OR but AND - some modules could be based on one style and other modules on another.