[nycbug-talk] Cogent and Sprint - a signal of things getting Oldschool?

Marc Spitzer mspitzer at gmail.com
Sat Nov 1 18:29:58 EDT 2008

On Sat, Nov 1, 2008 at 6:14 PM, Isaac Levy <ike at lesmuug.org> wrote:
> Charles and Alex,
> On Oct 31, 2008, at 8:13 PM, Alex Pilosov wrote:
>> On Fri, 31 Oct 2008, Charles Sprickman wrote:
>>> Nice.
>>> I've not followed the ups and downs of ISP dramas in a long time,
>>> but my
>>> gut feeling, even before reading the Renesys blog, was "oh, sprint
>>> still
>>> sells internets?".  Personally, I think this hurts Sprint the
>>> most.  My
>>> gut feeling is that they are something of a has-been in this market.
>> It's complicated. Despite being a "has-been", sprint maintains the "we
>> will not peer with you" reputation, and is the "hardest to establish
>> settlement-free peering" carrier. As a result, many people end up
>> using
>> only Sprint for transit (or, "the only transit we will admit to
>> having"),
>> so they *can* get other peering (it's a bit complicated - basically,
>> if
>> your transit is an existing peer, you won't get peering), in effect,
>> helping Sprint maintain this status.
>>> One thing that really has me wondering, and again, this is probably
>>> an
>>> Alex question, is an odd situation I ran into a few years back...
>>> I was
>>> toying around with two providers - L3 and HE.  I primarily wanted
>>> HE as
>>> backup, since L3 was not really soaking us and they generally have
>>> their
>>> shit together outside of the management/sales/install realms.  No
>>> matter
>>> how much I prepended our HE announcement, I just could not squash the
>>> inbound traffic.  Apparently HE buys transit from Cogent and there
>>> are a
>>> TON of people that shove all outbound traffic down a Cogent link if
>>> they
>>> have one.  This is not that much of a surprise (the volume of
>>> traffic was
>> *snicker* Yes, cogent is the "transit we use for outbound but we won't
>> admit to it".
>> The answer, of course, is not prepending it, but setting community
>> flags
>> telling HE to not announce this route to cogent, or to depreference
>> your
>> route while announcing to cogent, or some such. I don't know the
>> community
>> list for HE.
>>> though), but the thing that puzzled me when I ran a bunch of stuff
>>> through flow-tools was that I was seeing traffic from 1239 (Sprint)
>>> coming in through HE via Cogent.  I'm still puzzled as to what that
>>> was
>>> about - from my view, it looked like Sprint jamming traffic down
>>> Cogent
>>> rather than L3 (I'm certain Sprint and L3 peer).
>> It's complicated without looking at more details. You can't say
>> where it
>> *really* came from. What *could* be easily happening is that someone
>> (X)
>> only uses Sprint for inbound - so you see them behind Sprint. However,
>> X uses everyone else for outbound (including Cogent), who will
>> obviously
>> send it toward HE (paying customer).
> This thread is mind-blowingly interesting to me.
> This thread, and Miles' earlier email about 'hot potato' routing,
> makes me ask perhaps a stupid question:
> Why is routing not synchronous?  Why is sending more expensive than
> receiving packets- from a transit perspective?

Could you expand on what you mean by synchronous?


> Rocket-
> .ike
> _______________________________________________
> talk mailing list
> talk at lists.nycbug.org
> http://lists.nycbug.org/mailman/listinfo/talk

Freedom is nothing but a chance to be better.
Albert Camus

More information about the talk mailing list