When IP was in its infancy, network concerns were focused at Layers 2 (data link) and 3 (network) of the Open System Interconnection (OSI) model. As time went on, enhancements continued to these layers, and then the network industry ventured into Layer 4 (transport) for load balancing, Layer 5 (session) for VoIP QoS and Layer 6 (presentation) for encryption and/or acceleration. We now have reached Layer 7 (application). By definition, this layer was intended to provide network support services for application and end-user processes – addresses and QoS are identified; authentication, constraints and privacy security issues are addressed; network services for file transfers, e-mail and so on are invoked.
Now we’ve reached the point where network intelligence is pervasive across all layers of the OSI model. The IT industry has not been stagnant during this time frame and has moved forward into its own world of applications and end-user services as embodied in a service-oriented architecture (SOA). The SOA establishes a set of standards-based higher layers for IT services that apply equally to servers, storage and networks. In theory, the SOA creates layers on top of and interfacing with specific OSI services that exist within the network infrastructure. In practice, network service domains blur with IT domains, both lack inter-layer coordination and, in the eyes of some network vendors, actually can subsume or override the service layers of an IT architecture. This belief can become a major issue when the term “application aware” is taken to an extreme.
No one disputes that corporate networks must have the intelligence to be applications aware. What to do within the network using this applications knowledge is the issue. Deep packet-inspection techniques allow network components to know the resource destination, priority, QoS levels, data type, IT application, transaction type and content of a session flow, message or packet. What is not known is the transmission intent of the application. If the network attempts to reject, redirect or alter the message content, security status or service level, application QoS optimization and business process integrity might be compromised or lost.
Examples of the network intelligence problem abound. One obvious problem is transaction flow. Transactions are the bread and butter of the corporate IT world. The underpinning of transaction processing is the concept of workflow. Transactional workflow is predetermined and dependent upon the application usage in the corporate business process. The network cannot interfere in the IT orchestrated workflow without affecting the business process service-level requirements of the application. The next generation of IT transaction technology will be based upon grid services (for example, Globus 4.0) and will, by IT industry intent, become Web services, allowing the creation of dynamic transaction workflows. The intelligent network cannot independently optimize this real-time environment without intimate direction from the application.
The next problem area is virtualization. It is becoming an IT practice to virtualize compute and storage data center and distributed resources. The tools and products exist today to implement these IT technologies in a corporate environment. Although similar virtualization products exist for a network, their implementation is the exception rather than the rule. Without application awareness and IT control, network virtualization can override IT attempts to virtualize compute and storage resources.
Communications networking is at a critical juncture in its evolutionary cycle. Corporate decisions will need to be made as to where and in what form network intelligence will exist.
Too much independent network intelligence can create disaster unless it is dictatorially controlled and managed by IT SOA application processes. Too little independent network intelligence will negatively affect the successful implementation of a corporate business process-driven SOA.
