Mailbag: How intelligent should the network be?

In a recent newsletter we discussed Cisco’s stated intention to continually add more intelligence into the network. We also asked for your input as to whether or not you saw limits as to how much intelligence belongs in the network.

The most detailed input we have received so far came from David Green of NetGreen Consulting. Green observed that the placement of intelligence in the network tends to go in cycles, starting with a centralized approach in which all of the intelligence is under the control of a single entity, such as the way the phone companies used to control the PSTN. Over time, the approach typically changes and intelligence is driven to the edge of the network. Examples of distributed intelligence in the current WAN include the ability to tag packets for QoS purposes as well as security functionality such as firewalls and optimization functionality such as caching. Ideally, the approach of driving intelligence to the edge is standards based and is centrally managed and controlled.

Green was less concerned with where the intelligence resides and more concerned about the current lack of centralized management and control. He observed that in the current WAN environment a PC or a router can tag packets for special handling. However, there is no guarantee that the telco or ISP will provide anything other than best-effort service to these packets. He commented that while a few service providers are working to resolve this issue, most of the major service providers are busy absorbing their latest acquisition and attempting to integrate their disparate networks. It is reasonable to conclude that until these issues are resolved, the distributed intelligence will be less valuable than it could be.

We continue to seek your input on this topic. Let us know how much intelligence you think belongs in the network? Is there any specific intelligence that you believe should remain in the endpoints?