[191168] in North American Network Operators' Group
RE: Managed global low latency network with any to any connectivity
daemon@ATHENA.MIT.EDU (Naslund, Steve)
Wed Aug 24 12:01:39 2016
X-Original-To: nanog@nanog.org
From: "Naslund, Steve" <SNaslund@medline.com>
To: "nanog@nanog.org" <nanog@nanog.org>
Date: Wed, 24 Aug 2016 16:01:22 +0000
In-Reply-To: <CY4PR13MB09509EF2B2B37B0EC32E61A1E4EA0@CY4PR13MB0950.namprd13.prod.outlook.com>
Errors-To: nanog-bounces@nanog.org
The real issue in the request is that this person is looking for any-to-any=
connectivity which will require either a single L2 switching domain or a L=
3 routing domain. While waves, SDH, and SONET might be your layer one tran=
sport there are two major factors that are going to affect latency and jitt=
er the most. =20
1. Geography - Any point to any point has a minimum latency due to simple m=
ileage/medium constraints. You cannot possibly go any faster than the velo=
city of propagation over the media of your choice. For example, lowest lat=
ency at layer 1 would probably be P2P microwave (which has a faster velocit=
y of propagation than light over fiber) but that would not be an effective =
way to cross the Pacific ocean.
2. Routing/Switching queuing latency - If you want real any to any connecti=
vity you need routing or switching logic which takes time.
For example, lowest latency at layer 1 would probably be P2P microwave (whi=
ch has a faster velocity of propagation than fiber) but that would not be a=
n effective way to cross the Pacific ocean.
If you are doing an MPLS VPN architecture within the US, your routing/switc=
hing latency are probably going to be more significant than the layer one t=
echnology but when you go transoceanic your layer 1 latency becomes more si=
gnificant. The differences in electrical, free RF or optical (like microwa=
ve), and optical over fiber will vary by something like 30-40% of the speed=
of light over the mileage of the link. The routing/switching of an any-to=
-any architecture will probably dwarf most of the differences in media.
Steven Naslund
Chicago IL
-----Original Message-----
From: NANOG [mailto:nanog-bounces@nanog.org] On Behalf Of Rod Beck
Sent: Wednesday, August 24, 2016 10:45 AM
To: Ryan, Spencer; Arqam Gadit; nanog@nanog.org
Subject: Re: Managed global low latency network with any to any connectivit=
y
There are standard routes and there are low latency routes that serve mostl=
y traders. The latter charge a big premium. He said the lowest possible lat=
ency. That is a specialty market where the SLAs are in microseconds, not mi=
lliseconds. Many carriers have a division for ultra low latency. Hibernia A=
tlantic built express which is just used by financial traders. No one else =
can afford it. And since low latency is the name of the game, it means wave=
s or SDH or SONET. Not Ethernet switching.
Regards,
Roderick.
