On Dec 5, 2014, at 3:11 PM, Mike Horwath <[log in to unmask]> wrote:

> On Thu, Dec 04, 2014 at 09:08:02AM -0600, Andrew Hoyos wrote:
>
>> Virtual Chassis backplane is only 128Gbps (2x64Gbps), which seems
>> like alot, but given how much of traffic the exchange is doing,>
>> could quickly become an oversubscribed bottleneck.
>
> Or just move to 4550's at $9K each + cables for 256Gbps (2x128Gbps)
> stacking, and even add in the QSFP module for 40Gbps additional
> capacity.
>
> This is all resolvable with money - money from port fees would have
> covered this.
>
> Poking the dead-horse that really is alive and kicking.
>
> Leading from behind! I am Hindmost, back-seat driver, and overall
> captain obvious..

Before anyone starts talking about port fees,  more work needs to be done to come up with exactly what it is we need and then how to get it.

A few observations. From the traffic graph, and smoothing the anomalous burst caused by new connections, the growth rate probably matches the Internets minus something to account for the content players - which could be applied as anywhere from 50% to 100% YoY, market dependent.  If you're going to deploy additional hardware, how about a plan for at least three years of growth to minimize the cycles of ask/receive? A member survey asking about planning and growth would be helpful. We need a leaf node in another facility IMHO to also create some market pressure to keep costs low.

Best,

-M<