While application acceleration belongs in the network rack, it doesn't belong in packet-delivery devices, such as routers and firewalls, but rather in a new class of products focused on application delivery.

Network vendors have focused on elements such as IP addresses and TCP ports to provide a high-performance, secure packet-delivery infrastructure. But to ensure application-level performance and security, application delivery needs to focus on a different set of key elements - the users (who they are and their role), the application and the nature of their interaction (the session). For the future, application delivery will leverage, but remain separate from, packet delivery because of these differences in requirements and focus. It's like refrigerators and ovens - just because they're collocated doesn't mean they should be integrated.

Most application-performance problems occur because of inefficiencies and limitations within applications - not because of issues with TCP/IP. Application-performance problems stem mainly from chatty, WAN-inefficient application protocols stretched over long distances, and they cause long user wait time (latency).

Application-specific optimizations are the only way to significantly reduce latency. These protocol optimizations, as well as a comprehensive set of bandwidth-reduction capabilities and business-relevant application prioritization are possible only through understanding the user, application and session. Packet-delivery infrastructure understands none of these things. An application-delivery infrastructure has to.

The evolution of enterprise applications and networks are clarifying the differences between packet delivery and application delivery - and highlighting the need for separate infrastructures.