Re: routing issues to AWS via 2914(NTT)

[Bryan Socha <bryan () digitalocean com>]

I don't think anyone is blaming anyone, just trying to pass on information
where we see a problem.    We routed around it no problem.

Bryan Socha



On Fri, Jun 13, 2014 at 7:44 AM, Paul WALL <pauldotwall () gmail com> wrote:

> Amazon peers at many key exchanges, with dozens of hosting shops
> (where customers might share mutual infrastructure) like yours:
>
> https://www.peeringdb.com/view.php?asn=16509
>
> Rather than play the blame game with third-party transit providers,
> why not hit them up for some sessions?
>
> Drive Slow,
> Paul Wall
>
> On Fri, Jun 13, 2014 at 5:50 AM, Bryan Socha <bryan () digitalocean com>
> wrote:
> > Amazon hasn't reached out to us either...
> >
> > If you have other providers, use a combination of local-preference and
> the
> > customer communitiy strings with ntt to prepend around the circuit(s) in
> > nyc with the issue.  Just check your routing table, we found many going
> > through ntt to amazon and took awhile to get everything working as
> desired.
> > Bryan Socha
> > Network Engineer
> > DigitalOcean
> >
> > On Thu, Jun 12, 2014 at 12:55 PM, Christopher Rogers <phiber () phiber org>
> > wrote:
> >
> > > Could an IP engineer from AWS (16509/14618) and one from NTT (2914)
> kindly
> > > contact me off-list?  AS18888 is having some major reachability issues
> to
> > > you via 2914.  Several of our applications and users are reporting
> problems
> > > trying to reach various aws hosted services such as netflix and twilio.
> > >  I'm seeing almost 50% packet loss when transiting to you via 2914.
> > >  Forcing traffic onto 3356 clears the issue right up.  I've had to
> > > effectively shift all my ingress traffic off 2914 and de-pref aws
> as-path
> > > to force egress to other transit.
> > >
> > > We're a customer of 2914, but not AWS.  I've got a ticket open with
> 2914,
> > > and they've reached out to AWS, but it's been two days now and we
> haven't
> > > been getting any traction on this.
> > >
> > > thanks!
> > >
> > > -chris