On 10/04/2011 19:54, "Keith Mitchell" <[email protected]> wrote:
>I'm not sure it is so simple on this latter point. At present, there are >no trunk link technologies that scale cost-effectively above >10Gb/s - last I looked 40Gb/s gear costs 4x as much as 10Gb/s, and >100Gb/s is only just becoming available. This leads to major scaling >headaches at the core - the extensive use of inter-switch trunking of >multiple 10Gb/s links at major IXPs to handle traffic volumes of many >100s of Gb/s is just one painful consequence of this. I think IXP's are a rare and special case in this discussion. There are a lot of MPLS techniques for traffic engineering that work very well indeed, for example - using them to balance traffic on under-utilised trunks can save millions.
