Archived from groups: comp.dcom.lans.ethernet (
More info?)
Hansang Bae <uonr@alp.ee.pbz> writes:
>> Hansang Bae wrote:
>> > In datacenters, 10G or higher makes sense as people collapse more and
>> > more servers to a centralized super-servers. IT business is cyclical
>> > and the MF like functionality is coming back.
>In article <XrHec.23768$wP1.54064@attbi_s54>, gah@ugcs.caltech.edu says...
>> Within a datacenter, though it might still be a little fast for
>> most machines today. You can then run it through a switch
>> with gigabit to the clients.
>Not 10G for the servers, but for switch to switch connections. As you
>add more 1Gig capable servers to switches, they start to need bigger
>pipes to talk to the rest of the world. Our OSA (GigE) connections, as
>well as shared server segments (will very soon) need bigger up links.
>Gig Etherchannel and the like are becoming a hassle.
I can confirm that point of view / direction of development for our datacenter.
However, we already run fibre for switch-switch connections, and I don't see
a need to change this when we upgrade the interswitch links to 10GE. Once
we need to do that upgrade, the port density of GE aggregation switches
will be high enough to amortize the fibre cost over their ports, same as
it is now with our 100baseT aggregation.
To me copper 10GE can only make sense where existing wiring is CAT5 only,
and the cost to run fibre would be very high due to local constraints.
Maybe large buildings wired CAT5 only, where no more room to run new
cables between floors exists, and the old cables would need to be
replaced / disentangled? Dunno, in our vertical cabling between floors,
we ran fibre along with copper in the first place.
best regards
Patrick