That’s also true.. If you have a 10G connection between two DCs, and they can’t hash the traffic, you can only use 1/4th or 1/5th of the connection. Basically it is 10G but only 2G per flow. If you get transit at both places and then use a tunnel, which is a different service and may not satisfy all requirements, then you can use the full 10G, even with one flow. Otherwise you need to split it into 5 or more flows.
I guess people really don’t like Cogent judging by the fact that one unrelated email caused all this to happen again.. :-) > On 16 Oct 2018, at 18:01, David Hubbard <dhubb...@dino.hostasaurus.com> wrote: > > Yeah google is the issue for us. We provide web services and a LOT of our > customers have software that is making calls of various types to Google > services, or even just email delivery to Google hosted email; if all but a > Cogent transit link to a given data center were down, all of those customers’ > sites would begin failing at some level because the servers generally try v6 > if the application level wasn’t explicit. Cogent doesn’t seem to care since > their CEO is in some pissing match with Google. They must be deriving enough > revenue from last mile v4-only turn ups that they don’t really care about > dual stack customers. > > That being said, can’t say I’ve been impressed with their MPLS / metroE > offerings either. When doing the pricing/sizing routine on a project, I > learned that they have an internal concept of src-dst flows on those types of > circuits, and if they can’t see your labels, or otherwise hash the traffic, > or it all truly is point to point, you may not get the full bandwidth, or may > need to buy a capacity larger than what the flow will be. > > From: NANOG <nanog-boun...@nanog.org> on behalf of DaKnOb > <daknob....@gmail.com> > Date: Tuesday, October 16, 2018 at 10:06 AM > To: Dovid Bender <do...@telecurve.com> > Cc: NANOG <nanog@nanog.org> > Subject: Re: Whats going on at Cogent > > When I call and mention it I’m told that it’s HE’s fault (despite the lovely > cake), but when I also bring Google, then they tell me to get a different > provider just for this traffic, or meet them at an IX and send my traffic > from there. > > About the staff rotation I’ve seen it too, and I’ve also seen an increase in > salespeople calling, for example when an AS is registered etc. in addition to > the normal calls.. > > On 16 Oct 2018, at 16:54, Dovid Bender <do...@telecurve.com> wrote: > > They call me every few months. the last time they emailed me I said I wasn't > interested because of the HE issue. I have yet to get another email....... > > > On Tue, Oct 16, 2018 at 9:29 AM, Ca By <cb.li...@gmail.com> wrote: > > > On Tue, Oct 16, 2018 at 5:16 AM David Hubbard <dhubb...@dino.hostasaurus.com> > wrote: > Have had the same sales rep for several years now; unfortunately he has no > ability to fix their IPv6 peering issue so we’re slowly removing circuits, > but otherwise for a handful of 10gig DIA circuits it’s been stable. > > > Yep, this. Whenever Cogent calls, this is what i tell them. Black-holing HE > and Google ipv6 traffic, which is what they do if i use a default route from > them, is dead on arrival. Shows they make bad decisions and dont put the > customer first, or even create such an illusion. > > > From: NANOG <nanog-boun...@nanog.org> on behalf of Ryan Gelobter > <rya...@atwgpc.net> > Date: Tuesday, October 16, 2018 at 6:04 AM > To: NANOG <nanog@nanog.org> > Subject: Whats going on at Cogent > > Anyone else seen terrible support and high turnover of sales/account people > at Cogent the last few months? Is there something going on over there > internally? I'm sure some people will say Cogent has always been crap but in > the past their account reps and support were pretty good. It seems to have > gone downhill the last 12 months really bad. > > Regards, > Ryan >
