[38142] in North American Network Operators' Group
Re: QOS or more bandwidth
daemon@ATHENA.MIT.EDU (RJ Atkinson)
Tue May 29 13:20:42 2001
Message-Id: <5.1.0.14.2.20010529130927.00a70550@10.30.15.2>
Date: Tue, 29 May 2001 13:16:22 -0400
To: nanog@merit.edu
From: RJ Atkinson <rja@inet.org>
In-Reply-To: <20010529141406.736B1C790A@cesium.clock.org>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Errors-To: owner-nanog-outgoing@merit.edu
"E.B. Dreger" <eddy+public+spam@noc.everquick.net> asks:
| Although I generally agree, how does one keep QoS out of the core
| for CBR and jitter-sensitive applications?
Modern GigE/10GigE, properly implemented and deployed,
can provide more than adequately low jitter for voice use.
The ATM folks have their heads in the sand about this in
a huge way. Ethernet is generally cheaper than SONET. However,
modern WDM systems all have SONET interfaces and only some have
GigE interfaces -- so for WAN use SONET might still be the better
choice. In the LAN/MAN, GigE is a lot more cost effective approach,
speaking as someone who helped deploy a whole bunch of GigE-based
MANs in North America.
For folks who want extra fancy QoS mechanisms, for whatever
reason, some GigE vendors implement DiffServ in hardware with a full
8 queues/port and at least one vendor as an IP TDM approach that
is implemented in hardware. Custom ASIC hardware matters here
because it means that enabling QoS does NOT mean reducing the box's
packets/second forwarding rate.
Ran
rja@inet.org
