I hear a lot these days about ethernet deployment in metropolitan areas as access networks. Does this mean that there would be one big optical ethernet LAN (or MAN if you prefer) with fiber to the home connected by switches? In that case I have a few questions:
Wouldn't this lead to massive spanning trees in large cities? One bad configuration in the network would affect the whole network.
Will all IP traffic have to travel to the core even if it is destined for an intra-MAN destination? I cannot imagine that a ARP broadcast in a large MAN is feasable.
Will these networks co-exist with the current networks, or are they meant to replace the current infrastructure? In that case will VoIP replace the POTS? Will the MANs have sufficient QoS to handle voice?
I would appreciate any comments on these questions or your ideas on how the metro access networks of the future will look like.
No, it usually means using ethernet technology to employ e.g. MPLS instead of the classical SDH/ATM/MPOA approach. The scientific community in my city uses such a beast to connect the respective distributed campi of the participating organisations. (visit
in the UK there are fibre point to point links using ethernet (LAN extension service, or LES). The BT service is most common and it uses dedicated pair of fibres per link. You use it as a connection between 2 sites, up to 25 Km apart, although the costs can go up a lot over more than 5 to 10 Km.
other telcos either use the BT links, or their own Ethernet pipes or a mix to build more complex networks.
you can transport the ethernet directly on fibre, over SDH, using CWDM or DWDM to reduce the number of cores, over ATM etc. - all of these are in use here by various telcos. What you choose depends on how far you need to go, and how fibre "rich" the supplier is.
Some of these give you a strictly point to point connection, other are more flexible. so on some SDH gear you may have a point to multipoint style service, where 1 ethernet port at a central pointhas traffic on multiple VLANs, and packets from each VLAN may end up at a different port somewhere in the SDH cloud.
so, you may get your corporate internet feed on a 100M ethernet to the nearest Pop for your favorite ISP, or it may be an access line to your MPLS network and so on.
In that case I have a few questions:
you can do this, and some nets have used this in the UK - several school networks join sites in star or ring topology with ethernet layer 2 switches to glue it all together (usually to keep the price down).
even where this topology is in use, you normally get a limited group of users, usually part of the same community, so the layer 2 network is restricted in size.
worse yet - imagine a misconfigured set of routers which are replicating NetBIOS broadcasts by accident - or a loop replicating 100k broadcasts /sec :)
actually - you just described some of the drawbacks of layer 2 over MPLS - which is intended to give you an emulated Ethernet, possibly across a continent.
Nothing ever replaces all the installed base - these things will co-exist with conventional networks for a long time, even if they become the connection of choice.
And here there doesnt seem any prospect of universal broadband in the near future, never mind a fibre only service - so there should always be a need for something else. But - it may overtake some types of connection - anything over 2 Mbps looks to be a possibility.
In that case will VoIP
in the UK BT has announced they want all public PBXes to go Voip - but again it may take a while (the press release says 10 years - i suspect a lot longer to get your granny in a remote farmhouse to get rid of an analog phone)
Will the MANs have sufficient QoS to handle voice?
that bit is doable now.
But an classic ethernet point to point link doesnt need QoS - throw the bits in 1 end and they fall out the other a fixed time later in the same order. QoS only has an effect when you have 2 or more packets in a queue and you need to decide which to send 1st.
it is just the complicated electronics that link the ethernet pipes together that take account of QoS.
i just wish work would embrace it a bit more quickly - it would make my job designing networks easier.
it doesnt have to be worse - but it seems that a lot of the new "improved" WAN services are open to lots of problems that are only traditional on a LAN.
i also have a problem with treating a large net with relatively long latency WAN as a LAN without understanding what the side effects will be.
The more rope the easier to hang yourself :-) Nevertheless nobody forbids to do things the right way just because of more economic and flexible hardware. On the other hand I regret the lack of a native point-to-point ethernet standard ever since I built my first metro-ethernet back in '94.
Well I consider the latency issue a myth. A sensitive application will not run over a WAN either if not helpered, although I would consider it more of a problem if there existed a beast like metro-token-ring... :-)
Diego, there is a lot of work going on right now in different standard development organizations, to add features to Ethernet that make it better suited to the metropolitan networking world. IEEE Std 802.3ah ("Ethernet in the First Mile") defines a number of new PHYs and an OAM protocol for use in subscriber access networks. IEEE 802.1 is specifying provider bridges (using stacked VLANs) and end-to-end OAM for Ethernet, which will solve many of the issues that are associated with running the metro network as a LAN.
Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.