|
There is an article in EETimes by Fay Arjomandi of Vancouver-based Mobidia that may shake up the fans of the 10 year old stupid network principle. The stupid network essay calls for intelligence to reside at the edge of the network, rendering IP networks to plumbing pipes—with carriers ignorant of the application and services being transported.
The Arjomandi article suggests that what is really needed is a sharing of network intelligence and joint management of network efficiency practices. The article suggests that devices should be active, intelligent network elements within the network.
Such an approach would enable end-user devices to share the responsibilities of network management for uplinks while core network nodes manage the downlinks.
“This will require devices to be service-, network- and policy-aware. Such awareness will provide a much more comprehensive, efficient and wireless-friendly way to negotiate QoS with the network; enforce network policies; and queue, prioritize and manage traffic right at the source, based on its type, before sending data over precious network resources. This will eliminate the need for extra transactions over the wireless link for QoS parameter exchange. The device becomes a self-managing element within the network that manages uplink traffic.”
Interesting approach. Still, I suspect there are many IP network purists who would oppose any network-based means to manage demands on network capacity.
This article originally posted on the Telecom Trends weblog.
Sponsored byCSC
Sponsored byWhoisXML API
Sponsored byDNIB.com
Sponsored byRadix
Sponsored byVerisign
Sponsored byIPv4.Global
Sponsored byVerisign
I think some of these papers are so vague as to be almost useless.
I doubt consumer devices will ever be able to do more than say “this traffic is lower than average priority”, since otherwise it is open to abuse - witness Microsoft’s abuse of QoS bits in their applications - to try an obtain better than average service.
As such on the Internet side, the rules tend to be set in the router the ISP supplies (for leased lines), or the first piece of kit not under end user control.
Setting QoS bits in the first trustworthy device is precisely the “stupid network model”.
I think the paper suggest the same concept as cross-layer QoS… perhaps you should read more about this topic.. Cross-layer QoS is necessary for a successful QoS enforcement… beside the same model is followed by the diffserv where bits are set by the application… as well as IPv6… except that in these cases the intentional abuse or innocent mistake could cause drastic failure for the QoS enforcement over the network