As heterogeneous computing starts to grow, intelligent networking will be the facilitators of smart enterprise systems architecture. Basically hardware vendors are beginning to put intelligent silicon on network adaptors. This provides the ability through deep packet inspection to realistically provide Network Function Virtualization (NFV) and true Software Defined Networks (SDN) as a part of hardware/software computing infrastructure. This requires an intelligent NIC, Software Defined Networks (SDN) & Web Services/Cloud Servers must be engineered to “be aware” of the intelligence in the hardware so that software can make smart choices based on business logic context.
See more information about this trend here:
The IaaS and PaaS cloud models allow architects to decouple components of an application or enterprise system into the lowest functional components and design for failure how these pieces can be utilized as “independent black boxes” to form an application. This allows for provisioning elasticity and resiliency of individual components and their states in the inevitable event of hardware or software failure.
One of the least understood impacts of this approach is that the message queues used by components can become the most important elements in assuring availability, scalability and ultimate reliability. In essence the messaging infrastructure components become the most critical parts of an applications infrastructure designed to exploit elasticity. If you envision these Enterprise Apps as complex organisms, then the message queues and their reliability become mission critical organs of the living, agile enterprise architecture. Components such as controller apps, databases and such should be isolated allowing buffering of request along with replies making the network of components more durable and state independent facilitating failover and scalability.