Print

Print


As a transport seller, We're seeing most networks most commonly buy
linear paths to two or more IXes. Typically going opposite directions
from their home base. We're also starting to see smaller operators
exclusively doing remote peering. A typical set up is buying identical
transport to the 2 adjacent IX buildings.

On wavelengths, the link failure can be passed back to from end to
end. Granted, a long-haul wavelength isn't as stable as a cross
connect.

If a network doesn't have any remote gear the convergence is governed
by EBGP vs IBGP/IGP if there is an additional router. For remote
setups, in particular with Ethernet transport we're seeing more and
more BFD on the sessions.

Jay

On Mon, Oct 30, 2017 at 8:27 AM, Michael Hare <[log in to unmask]> wrote:
> For those buying wavelengths (compared to an mpls circuit), are you buying diverse transport?  If not, how are you dealing with convergence stability?
>
> -Michael
>
>>>-----Original Message-----
>>>From: MICE Discuss [mailto:[log in to unmask]] On Behalf
>>>Of Colin Baker
>>>Sent: Friday, October 27, 2017 3:29 PM
>>>To: [log in to unmask]
>>>Subject: Re: [MICE-DISCUSS] Multi-IX Traffic Management?
>>>
>>>On 27.10.2017 15:01, Matthew Beckwell wrote:
>>>> For my own curiosity:
>>>>
>>>> I've seen quite a few networks in the last year or so connecting
>>>> themselves to other out-of-market IX's (I presume getting some cheap
>>>> wavelengths and bringing it back to Minnesota).
>>>>
>>>> A couple of questions for those of you that are doing this:
>>>>
>>>> 1. I'm assuming this is enabling you to accommodate requirements from
>>>> other networks that require peering at multiple locations, even though
>>>> technically your network doesn't really extend that far (or maybe it's
>>>> something else?).
>>>
>>>For us, the goal is to reach networks which are not on exchanges which
>>>we're already connected.  The big ones are at MICE, but there are other
>>>exchanges we can get to reasonably inexpensively that have such a large
>>>number of networks connected that it makes sense financially (compared
>>>to buying more transit).  Traffic from a huge number of ASNs,
>>>particularly if most are on a route server, can really add up.
>>>
>>>Peering in multiple locations is a nice benefit to this, but usually not
>>>a primary goal for us.
>>>
>>>> 2. How are folks monitoring, managing, or manipulating this
>>>> out-of-market traffic? (Using BGP MED, Prepending, Localpref, or some
>>>> other mechanism to "prefer" your traffic enter and exit closer to
>>>> where you really are?)
>>>
>>>If you haven't already, install as-stats
>>>(https://github.com/manuelkasper/AS-Stats) to monitor things, which will
>>>make it clear if something is out of whack.  In most cases, our IGP
>>>handles outbound traffic in a way the makes sense.  Inbound traffic can
>>>be trickier - a few networks will publish action communities or accept
>>>MEDs, and we utilize those tools when available.  Prepending is
>>>occasionally useful.  Announcing more specifics in certain spots always
>>>felt like a jerk move so I tend to avoid that, but that's a thing I
>>>suppose.
>>>
>>>--
>>>Colin Baker
>>>SupraNet Communications, Inc.
>>>(608) 572-7634
>>>[log in to unmask]
>>>
>>>This message is subject to the SupraNet Email Confidentiality Policy
>>>which is located at http://supranet.net/confidentiality



-- 
Jay Hanke
CTO
Neutral Path Communications
3 Civic Center Plaza, Suite 204
Mankato, MN 56001
(507) 327-2398 mobile
[log in to unmask]
www.neutralpath.net