Ethernet over ADSL: Difference between revisions

From AAISP Support Site
(random idea)
 
m (add note)
Line 1: Line 1:
Note: This isn’t an official AAISP thing, but something I thought about relating to [http://revk.www.me.uk/2011/07/how-bt-should-do-it.html How BT should do it].
Note: This isn’t an official AAISP thing, and may be wrong, but something I thought about relating to [http://revk.www.me.uk/2011/07/how-bt-should-do-it.html How BT should do it].


Hopefully there will be other ideas as well on how to do this…
Hopefully there will be other ideas as well on how to do this…

Revision as of 23:01, 13 October 2011

Note: This isn’t an official AAISP thing, and may be wrong, but something I thought about relating to How BT should do it.

Hopefully there will be other ideas as well on how to do this…

Michael 00:00, 14 October 2011 (BST)

Ethernet via ADSL Ideas

The CPE widely used for ADSL will typically have routing functionality,

This may be considered excessive as the device often has only 2 routes, one to the Internet and one to the user's network.

As a result of this, there is a tradeoff to be made regarding functionality, typically a DNS relay and DHCP server will be included at the very least.

ISP hosted virtual routers

It is possible to consider moving the routing task to the ISP, and give each subscriber their own virtual router running on the ISP’s hardware.

The CPE would have one or more ethernet ports and generally deliver raw frames between these ports and the ISP, and also tell the ISP if the ethernet jacks function or not.

CPE Features

Generally only features that either require support locally would be offered, such as Power over Ethernet (source or sink) and Synchronous Ethernet, perhaps driven from the national ISDN master timing sources.

Also, if this CPE is mounted externally, the ISP can repair it without incurring a failed visit. VLAN services can provide access to multiple routers, multiple ISP or even direct access to an Internet exchange peering LAN from home. As long as the user is not required to run PPPoE then the service is stateless and there are no obscure authentication problems to be had.

This does not preclude the subscriber connecting their own firewall or router between the CPE and their other equipment, and arranging for several blocks of address space to be routed to it for assignment deeper in their network, and can work well.

Advantages

  • Energy efficiency - by centralising routing services, users can have a greener CPE. And it could be better for the environment as the energy savings from all CPE added together may not even be reached by a loaded ISP router.
  • Less network hops - lower latency.
  • Features - want per MAC address accounting, uptime graphs, sophisticated firewalling? These could be offerable as a managed service and can be kept up to date.
  • Reliability - by making CPE simpler, crashes could be less likely. The ISP manages their datacentre equipment and can restart it as needed without visiting any customers. Also possibly good for subscribers connecting ethernet kiosks or ledboards back to their hq.
  • IPv4 conservation - the virtual routers can collaborate on address space so that addresses are issued as needed from a single pool or allocated statically as agreed between ISP and customer.

Disadvantages

  • Privacy - the ISP, and anyone eavesdropping on the WAN circuit, will typically be able to see more info about what the subscriber equipment is doing. Those that care about this are likely to want their own equipment to protect it however.
  • Packet header overhead - adding Ethernet frames will decrease the payload available.
  • WAN loss effects more noticable - some subscribers may consider this an advantage.