NANOG Digest, Vol 94, Issue 4

Arthur Liew arthurliew80 at gmail.com
Fri Nov 6 12:57:16 UTC 2015


Hi Brandon,

Does Border6 works for inbound traffic engineering too?

We were exploring inbound traffic engineering a while ago, and had attended
Huawei RR+ demo. The idea is great, but they only support their own router
netflow currently

We are still looking for options.It'll be great if you can share more if it
supports.

Rgds
Arthur Liew
(CCIE#38181)

On Fri, Nov 6, 2015 at 8:00 PM, <nanog-request at nanog.org> wrote:

> Send NANOG mailing list submissions to
>         nanog at nanog.org
>
> To subscribe or unsubscribe via the World Wide Web, visit
>         http://mailman.nanog.org/mailman/listinfo/nanog
> or, via email, send a message with subject or body 'help' to
>         nanog-request at nanog.org
>
> You can reach the person managing the list at
>         nanog-owner at nanog.org
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of NANOG digest..."
>
>
> Today's Topics:
>
>    1. Re: Internap route optimization (chip)
>    2. Re: Internap route optimization (Mike Hammett)
>    3. Looking for a comcast NOC and/or peering contact (Eric Sieg)
>    4. Re: Internap route optimization (Sebastian Spies)
>    5. Re: Internap route optimization (Fred Hollis)
>    6. RE: Internap route optimization (Eric Van Tol)
>    7. Re: AT&T Wholesale (Alex Forster)
>    8. Long-haul 100Mbps EPL circuit throughput issue (Eric Dugas)
>    9. Re: Long-haul 100Mbps EPL circuit throughput issue (alvin nanog)
>   10. Re: Long-haul 100Mbps EPL circuit throughput issue (Bob Evans)
>   11. Re: Long-haul 100Mbps EPL circuit throughput issue (Pablo Lucena)
>   12. Re: Long-haul 100Mbps EPL circuit throughput issue (Pablo Lucena)
>   13. Re: Long-haul 100Mbps EPL circuit throughput issue
>       (Theodore Baschak)
>   14. Re: Long-haul 100Mbps EPL circuit throughput issue (Greg Foletta)
>   15. Re: Long-haul 100Mbps EPL circuit throughput issue
>       (William Herrin)
>   16. Re: Long-haul 100Mbps EPL circuit throughput issue (Pablo Lucena)
>   17. Re: Internap route optimization (Brandon Wade)
>   18. Youtube CDN unreachable over IPv6 (Seth Mos)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Thu, 5 Nov 2015 07:18:08 -0500
> From: chip <chip.gwyn at gmail.com>
> To: Christopher Morrow <morrowc.lists at gmail.com>
> Cc: Fred Hollis <fred at web2objects.com>, nanog list <nanog at nanog.org>
> Subject: Re: Internap route optimization
> Message-ID:
>         <
> CABGzhdu8t8HW8RC4Tk+XWtrEtJZZuia_DcN-A2E_7TUv-PBGiQ at mail.gmail.com>
> Content-Type: text/plain; charset=UTF-8
>
> Just to be clear, Internap's solution doesn't use "more specifics" to steer
> traffic.  The mechanisms in place to protect yourself from normal route
> leaking should apply just the same.
>
> --chip
>
>
> On Thu, Nov 5, 2015 at 5:01 AM, Christopher Morrow <
> morrowc.lists at gmail.com>
> wrote:
>
> > Also, please, if you use one of this sort of device filter your
> > prefixes toward your customers/peers/transits... Do not be the next
> > person to leak their internap-box-routes to the world, m'kay? :)
> >
> > On Thu, Nov 5, 2015 at 8:53 PM, Fred Hollis <fred at web2objects.com>
> wrote:
> > > Hi,
> > >
> > > No particular experience with Internaps optimization...however I
> > wouldn't be
> > > that sure if I would use it within our networks, because you always
> have
> > the
> > > conflict of this not being their core business as they want to sell
> their
> > > optimized IP transit.
> > >
> > > However, some time ago we tried Border6 in an evaluation and then
> finally
> > > put it into production. Not only the optimization is nice, but the
> > reporting
> > > is so extremely detailed making it very transparent where the transit
> has
> > > congestion issues and which prefix is routed (in and out) through which
> > > upstream.
> > >
> > > For sure traffic engineering/optimization is not a trivial task but
> > requires
> > > deep thinking and understanding of the whole BGP and routing picture.
> > >
> > >
> > > On 05.11.2015 at 09:03 Paras wrote:
> > >>
> > >> Does anyone know or have any experience with Internap's route
> > >> optimization? Is it any good?
> > >>
> > >> I've heard of competing solutions as well, such as the one provided by
> > >> Noction.
> > >>
> > >> Thanks for your input,
> > >> Paras
> > >>
> > >
> >
>
>
>
> --
> Just my $.02, your mileage may vary,  batteries not included, etc....
>
>
> ------------------------------
>
> Message: 2
> Date: Thu, 5 Nov 2015 07:21:05 -0600 (CST)
> From: Mike Hammett <nanog at ics-il.net>
> Cc: nanog at nanog.org
> Subject: Re: Internap route optimization
> Message-ID:
>         <962573932.10921.1446729692705.JavaMail.mhammett at ThunderFuck>
> Content-Type: text/plain; charset=utf-8
>
> Keep in mind that most do not optimize inbound traffic, only outbound.
>
>
>
>
> -----
> Mike Hammett
> Intelligent Computing Solutions
> http://www.ics-il.com
>
> ----- Original Message -----
>
> From: "Paras" <paras at protrafsolutions.com>
> To: nanog at nanog.org
> Sent: Thursday, November 5, 2015 2:03:41 AM
> Subject: Internap route optimization
>
> Does anyone know or have any experience with Internap's route
> optimization? Is it any good?
>
> I've heard of competing solutions as well, such as the one provided by
> Noction.
>
> Thanks for your input,
> Paras
>
>
>
>
> ------------------------------
>
> Message: 3
> Date: Thu, 5 Nov 2015 08:23:03 -0500
> From: Eric Sieg <eric.sieg at gmail.com>
> To: nanog at nanog.org
> Subject: Looking for a comcast NOC and/or peering contact
> Message-ID:
>         <
> CAEcHXr6EXg2M2uc-JkuSG3EjTPU4nra-sRiUhAR0fDyKHLUKdg at mail.gmail.com>
> Content-Type: text/plain; charset=UTF-8
>
> If you could contact me off-list, it would be greatly appreciated!
>
> -Eric
>
>
> ------------------------------
>
> Message: 4
> Date: Thu, 5 Nov 2015 14:56:10 +0100
> From: Sebastian Spies <s+Mailinglisten.nanog at sloc.de>
> To: nanog at nanog.org
> Subject: Re: Internap route optimization
> Message-ID: <563B5FFA.8020409 at sloc.de>
> Content-Type: text/plain; charset=utf-8
>
> Hey Mike,
>
> do you know route optimizers that actually do optimize inbound traffic?
> We, at datapath.io, are currently working on this and could not find
> another one that does it.
>
> Best,
> Sebastian
>
> Am 05.11.2015 um 14:21 schrieb Mike Hammett:
> > Keep in mind that most do not optimize inbound traffic, only outbound.
> >
> >
> >
> >
> > -----
> > Mike Hammett
> > Intelligent Computing Solutions
> > http://www.ics-il.com
> >
> > ----- Original Message -----
> >
> > From: "Paras" <paras at protrafsolutions.com>
> > To: nanog at nanog.org
> > Sent: Thursday, November 5, 2015 2:03:41 AM
> > Subject: Internap route optimization
> >
> > Does anyone know or have any experience with Internap's route
> > optimization? Is it any good?
> >
> > I've heard of competing solutions as well, such as the one provided by
> > Noction.
> >
> > Thanks for your input,
> > Paras
> >
> >
>
>
> ------------------------------
>
> Message: 5
> Date: Thu, 5 Nov 2015 15:00:53 +0100
> From: Fred Hollis <fred at web2objects.com>
> To: nanog at nanog.org
> Subject: Re: Internap route optimization
> Message-ID: <563B6115.808 at web2objects.com>
> Content-Type: text/plain; charset=utf-8; format=flowed
>
> Border6 offers such an option based on prependings and bgp communities.
> But honestly, I didn't test that feature yet. And I'm not that sure how
> much sense it makes since it probably requires quite a lot global BGP
> updates... makes routers even more busy than they are currently.
>
> On 05.11.2015 at 14:56 Sebastian Spies wrote:
> > Hey Mike,
> >
> > do you know route optimizers that actually do optimize inbound traffic?
> > We, at datapath.io, are currently working on this and could not find
> > another one that does it.
> >
> > Best,
> > Sebastian
> >
> > Am 05.11.2015 um 14:21 schrieb Mike Hammett:
> >> Keep in mind that most do not optimize inbound traffic, only outbound.
> >>
> >>
> >>
> >>
> >> -----
> >> Mike Hammett
> >> Intelligent Computing Solutions
> >> http://www.ics-il.com
> >>
> >> ----- Original Message -----
> >>
> >> From: "Paras" <paras at protrafsolutions.com>
> >> To: nanog at nanog.org
> >> Sent: Thursday, November 5, 2015 2:03:41 AM
> >> Subject: Internap route optimization
> >>
> >> Does anyone know or have any experience with Internap's route
> >> optimization? Is it any good?
> >>
> >> I've heard of competing solutions as well, such as the one provided by
> >> Noction.
> >>
> >> Thanks for your input,
> >> Paras
> >>
> >>
>
>
> ------------------------------
>
> Message: 6
> Date: Thu, 5 Nov 2015 06:17:33 -0500
> From: Eric Van Tol <eric at atlantech.net>
> To: NANOG <nanog at nanog.org>
> Subject: RE: Internap route optimization
> Message-ID:
>         <2C05E949E19A9146AF7BDF9D44085B86720A812942 at exchange.aoihq.local>
> Content-Type: text/plain; charset="utf-8"
>
> TL;DR: Not worth it unless you have only a few transit providers and are a
> content-heavy network with little inbound traffic.
>
> We used the Internap FCP for a long time (10 or so years). In general, we
> were satisfied with it, but honestly, after not having it in our network
> for the past year and a half, we really don't notice a difference. We
> primarily purchased it to keep transit costs down, but as we kept boosting
> our minimums with providers, it became less and less about transit costs
> and more about performance.
>
> Boxes like these really work best if your network is a content-heavy
> network (more outbound than inbound). Sure, it will route around poorly
> performing paths, but IMO it's not worth the money and yearly maintenance
> fees just for this. I always said that it must be doing a good job since we
> never got complaints about packet loss in an upstream network, but now that
> the device is gone, we still don't get complaints about packet loss in an
> upstream's network. :-/
>
> The biggest problem that we found was that it just was not actively
> developed (at the time, not sure about now). New software features were
> non-existent for years. Bugs were not fixed in a timely manner. Given what
> we were paying in yearly maintenance fees, it just wasn't worth it to keep
> around. It also wasn't scalable as we kept adding more transit interfaces,
> given that there were a fixed amount of capture ports. Adding non-transit
> peering into the mix was also complicated and messed with the route
> decision algorithms. Maybe things have changed.
>
> As far as technicals, it seemed to work fine. One of the really only
> annoying things about it were remote users who think that a UDP packet
> hitting their firewall from its automatic traceroute mechanism were 'DDoS'
> and threats of lawyers/the wrath of god almighty would come down upon us
> for sending unauthorized packets to their precious and delicate network.
> You would definitely also want to make sure that you filter announcements
> so you don't accidentally start sending longer paths to your upstreams or
> customer peers, but if you run BGP, you already do that, amirite?!
>
> -evt
>
> > -----Original Message-----
> > From: NANOG [mailto:nanog-bounces at nanog.org] On Behalf Of Paras
> > Sent: Thursday, November 05, 2015 3:04 AM
> > To: nanog at nanog.org
> > Subject: Internap route optimization
> >
> > Does anyone know or have any experience with Internap's route
> > optimization? Is it any good?
> >
> > I've heard of competing solutions as well, such as the one provided by
> > Noction.
> >
> > Thanks for your input,
> > Paras
>
>
> ------------------------------
>
> Message: 7
> Date: Thu, 5 Nov 2015 02:38:44 +0000
> From: Alex Forster <alex at alexforster.com>
> To: "nanog at nanog.org" <nanog at nanog.org>
> Subject: Re: AT&T Wholesale
> Message-ID: <D2602A95.2DAA%alex at alexforster.com>
> Content-Type: text/plain; charset="iso-8859-1"
>
> Actually, I'd appreciate pointers to a good rep with AT&T or Verizon for
> Layer 2 services as well ? no preference for which region of the country.
> Thanks to anyone who can help!
>
> Alex Forster
>
>
>
> On 11/4/15, 12:53 PM, "NANOG on behalf of Sam Norris"
> <nanog-bounces at nanog.org on behalf of Sam at SanDiegoBroadband.com> wrote:
>
> >Hey everyone,
> >
> >Can someone send me privately the contact info for an AT&T Wholesale rep
> >for
> >Metro E / VPLS / Layer 2 stuff here in the SouthWest region?  Their
> >website is
> >not very informative on how to make any contact with the wholesale group.
> >
> >Thx,
> >Sam
> >
>
>
>
> ------------------------------
>
> Message: 8
> Date: Thu, 5 Nov 2015 16:48:51 -0500
> From: Eric Dugas <edugas at unknowndevice.ca>
> To: nanog at nanog.org
> Subject: Long-haul 100Mbps EPL circuit throughput issue
> Message-ID:
>         <CALKrK4na=wKSi=
> vf4EgPetoXJxaHTxfGt8HXYr6CE+xPk1Vy4g at mail.gmail.com>
> Content-Type: text/plain; charset=UTF-8
>
> Hello NANOG,
>
> We've been dealing with an interesting throughput issue with one of our
> carrier. Specs and topology:
>
> 100Mbps EPL, fiber from a national carrier. We do MPLS to the CPE providing
> a VRF circuit to our customer back to our data center through our MPLS
> network. Circuit has 75 ms of latency since it's around 5000km.
>
> Linux test machine in customer's VRF <-> SRX100 <-> Carrier CPE (Cisco
> 2960G) <-> Carrier's MPLS network <-> NNI - MX80 <-> Our MPLS network <->
> Terminating edge - MX80 <-> Distribution switch - EX3300 <-> Linux test
> machine in customer's VRF
>
> We can full the link in UDP traffic with iperf but with TCP, we can reach
> 80-90% and then the traffic drops to 50% and slowly increase up to 90%.
>
> Any one have dealt with this kind of problem in the past? We've tested by
> forcing ports to 100-FD at both ends, policing the circuit on our side,
> called the carrier and escalated to L2/L3 support. They tried to also
> police the circuit but as far as I know, they didn't modify anything else.
> I've told our support to make them look for underrun errors on their Cisco
> switch and they can see some. They're pretty much in the same boat as us
> and they're not sure where to look at.
>
> Thanks
> Eric
>
>
> ------------------------------
>
> Message: 9
> Date: Thu, 5 Nov 2015 15:19:12 -0800
> From: alvin nanog <nanogml at Mail.DDoS-Mitigator.net>
> To: Eric Dugas <edugas at unknowndevice.ca>
> Cc: nanog at nanog.org
> Subject: Re: Long-haul 100Mbps EPL circuit throughput issue
> Message-ID: <20151105231912.GA17090 at Mail.DDoS-Mitigator.net>
> Content-Type: text/plain; charset=us-ascii
>
>
> hi eric
>
> On 11/05/15 at 04:48pm, Eric Dugas wrote:
> ...
> > Linux test machine in customer's VRF <-> SRX100 <-> Carrier CPE (Cisco
> > 2960G) <-> Carrier's MPLS network <-> NNI - MX80 <-> Our MPLS network <->
> > Terminating edge - MX80 <-> Distribution switch - EX3300 <-> Linux test
> > machine in customer's VRF
> >
> > We can full the link in UDP traffic with iperf but with TCP, we can reach
> > 80-90% and then the traffic drops to 50% and slowly increase up to 90%.
>
> if i was involved with these tests, i'd start looking for "not enough tcp
> send
> and tcp receive buffers"
>
> for flooding at 100Mbit/s, you'd need about 12MB buffers ...
>
> udp does NOT care too much about dropped data due to the buffers,
> but tcp cares about "not enough buffers" .. somebody resend packet#
> 1357902456 :-)
>
> at least double or triple the buffers needed to compensate for all kinds of
> network whackyness:
> data in transit, misconfigured hardware-in-the-path, misconfigured iperfs,
> misconfigured kernels, interrupt handing, etc, etc
>
> - how many "iperf flows" are you also running ??
>         - running dozen's or 100's of them does affect thruput too
>
> - does the same thing happen with socat ??
>
> - if iperf and socat agree with network thruput, it's the hw somewhere
>
> - slowly increasing thruput doesn't make sense to me ... it sounds like
> something is cacheing some of the data
>
> magic pixie dust
> alvin
>
> > Any one have dealt with this kind of problem in the past? We've tested by
> > forcing ports to 100-FD at both ends, policing the circuit on our side,
> > called the carrier and escalated to L2/L3 support. They tried to also
> > police the circuit but as far as I know, they didn't modify anything
> else.
> > I've told our support to make them look for underrun errors on their
> Cisco
> > switch and they can see some. They're pretty much in the same boat as us
> > and they're not sure where to look at.
> >
>
>
> ------------------------------
>
> Message: 10
> Date: Thu, 5 Nov 2015 15:31:39 -0800
> From: "Bob Evans" <bob at FiberInternetCenter.com>
> To: "Eric Dugas" <edugas at unknowndevice.ca>
> Cc: nanog at nanog.org
> Subject: Re: Long-haul 100Mbps EPL circuit throughput issue
> Message-ID: <2af1f897aece8062f5744396fed1559a.squirrel at 66.201.44.180>
> Content-Type: text/plain;charset=iso-8859-1
>
> Eric,
>
> I have seen that happen.
>
> 1st double check that the gear is truly full duplex....seems like it may
> claim it is and you just discovered it is not. That's always been an issue
> with manufactures claiming they are full duplex and on short distances
> it's not so noticeable.
>
> Try to perf in both directions at the same time and it become obvious.
>
> Thank You
> Bob Evans
> CTO
>
>
>
>
> > Hello NANOG,
> >
> > We've been dealing with an interesting throughput issue with one of our
> > carrier. Specs and topology:
> >
> > 100Mbps EPL, fiber from a national carrier. We do MPLS to the CPE
> > providing
> > a VRF circuit to our customer back to our data center through our MPLS
> > network. Circuit has 75 ms of latency since it's around 5000km.
> >
> > Linux test machine in customer's VRF <-> SRX100 <-> Carrier CPE (Cisco
> > 2960G) <-> Carrier's MPLS network <-> NNI - MX80 <-> Our MPLS network <->
> > Terminating edge - MX80 <-> Distribution switch - EX3300 <-> Linux test
> > machine in customer's VRF
> >
> > We can full the link in UDP traffic with iperf but with TCP, we can reach
> > 80-90% and then the traffic drops to 50% and slowly increase up to 90%.
> >
> > Any one have dealt with this kind of problem in the past? We've tested by
> > forcing ports to 100-FD at both ends, policing the circuit on our side,
> > called the carrier and escalated to L2/L3 support. They tried to also
> > police the circuit but as far as I know, they didn't modify anything
> else.
> > I've told our support to make them look for underrun errors on their
> Cisco
> > switch and they can see some. They're pretty much in the same boat as us
> > and they're not sure where to look at.
> >
> > Thanks
> > Eric
> >
>
>
>
>
> ------------------------------
>
> Message: 11
> Date: Thu, 5 Nov 2015 21:17:01 -0500
> From: Pablo Lucena <plucena at coopergeneral.com>
> To: bob at fiberinternetcenter.com
> Cc: Eric Dugas <edugas at unknowndevice.ca>, "NANOG Operators' Group"
>         <nanog at nanog.org>
> Subject: Re: Long-haul 100Mbps EPL circuit throughput issue
> Message-ID:
>         <CAH+2GVLSUYMo-XGdQc1knYEq1hA13=NOQf4jP=
> 2fiNFMQoC5aw at mail.gmail.com>
> Content-Type: text/plain; charset=UTF-8
>
> With default window size of 64KB, and a delay of 75 msec, you should only
> get around 7Mbps of throughput with TCP.
>
> You would need a window size of about 1MB in order to fill up the 100 Mbps
> link.
>
> 1/0.75 = 13.333 (how many RTTs in a second)
> 13.333 * 65535 * 8 = 6,990,225.24 (about 7Mbps)
>
> You would need to increase the window to 1,048,560 KB, in order to get
> around 100Mbps.
>
> 13.333 * 1,048,560 * 8 = 111,843,603.84 (about 100 Mbps)
>
>
> *Pablo Lucena*
>
> *Cooper General Global Services*
>
> *Network Administrator*
>
> *Office: 305-418-4440 ext. 130*
>
> *plucena at coopergeneral.com <plucena at coopergeneral.com>*
>
> On Thu, Nov 5, 2015 at 6:31 PM, Bob Evans <bob at fiberinternetcenter.com>
> wrote:
>
> > Eric,
> >
> > I have seen that happen.
> >
> > 1st double check that the gear is truly full duplex....seems like it may
> > claim it is and you just discovered it is not. That's always been an
> issue
> > with manufactures claiming they are full duplex and on short distances
> > it's not so noticeable.
> >
> > Try to perf in both directions at the same time and it become obvious.
> >
> > Thank You
> > Bob Evans
> > CTO
> >
> >
> >
> >
> > > Hello NANOG,
> > >
> > > We've been dealing with an interesting throughput issue with one of our
> > > carrier. Specs and topology:
> > >
> > > 100Mbps EPL, fiber from a national carrier. We do MPLS to the CPE
> > > providing
> > > a VRF circuit to our customer back to our data center through our MPLS
> > > network. Circuit has 75 ms of latency since it's around 5000km.
> > >
> > > Linux test machine in customer's VRF <-> SRX100 <-> Carrier CPE (Cisco
> > > 2960G) <-> Carrier's MPLS network <-> NNI - MX80 <-> Our MPLS network
> <->
> > > Terminating edge - MX80 <-> Distribution switch - EX3300 <-> Linux test
> > > machine in customer's VRF
> > >
> > > We can full the link in UDP traffic with iperf but with TCP, we can
> reach
> > > 80-90% and then the traffic drops to 50% and slowly increase up to 90%.
> > >
> > > Any one have dealt with this kind of problem in the past? We've tested
> by
> > > forcing ports to 100-FD at both ends, policing the circuit on our side,
> > > called the carrier and escalated to L2/L3 support. They tried to also
> > > police the circuit but as far as I know, they didn't modify anything
> > else.
> > > I've told our support to make them look for underrun errors on their
> > Cisco
> > > switch and they can see some. They're pretty much in the same boat as
> us
> > > and they're not sure where to look at.
> > >
> > > Thanks
> > > Eric
> > >
> >
> >
> >
>
>
> ------------------------------
>
> Message: 12
> Date: Thu, 5 Nov 2015 21:18:38 -0500
> From: Pablo Lucena <plucena at coopergeneral.com>
> To: bob <bob at fiberinternetcenter.com>
> Cc: Eric Dugas <edugas at unknowndevice.ca>, "NANOG Operators' Group"
>         <nanog at nanog.org>
> Subject: Re: Long-haul 100Mbps EPL circuit throughput issue
> Message-ID:
>         <
> CAH+2GV+8kO1LEiCWri2NYyB5GsUbsc02uSTmVjK5+_AQgcVQqQ at mail.gmail.com>
> Content-Type: text/plain; charset=UTF-8
>
> > With default window size of 64KB, and a delay of 75 msec, you should only
> > get around 7Mbps of throughput with TCP.
> >
> > You would need a window size of about 1MB in order to fill up the 100
> Mbps
> > link.
> >
> > 1/0.75 = 13.333 (how many RTTs in a second)
> > 13.333 * 65535 * 8 = 6,990,225.24 (about 7Mbps)
> >
> > You would need to increase the window to 1,048,560 KB, in order to get
> > around 100Mbps.
> >
> > 13.333 * 1,048,560 * 8 = 111,843,603.84 (about 100 Mbps)
> >
> >>
> >>
> >>
> >
> ?I realized I made a typo:
>
> 1/*0.075* = 13.333
>
> not
>
> 1/0.75 = 13.333
>
>
> ?
>
>
> ------------------------------
>
> Message: 13
> Date: Thu, 5 Nov 2015 20:27:27 -0600
> From: Theodore Baschak <theodore at ciscodude.net>
> To: NANOG Operators' Group <nanog at nanog.org>
> Subject: Re: Long-haul 100Mbps EPL circuit throughput issue
> Message-ID: <4A6A5425-ADA1-49D6-89D5-DE189A53D8FD at ciscodude.net>
> Content-Type: text/plain;       charset=utf-8
>
> > On Nov 5, 2015, at 8:18 PM, Pablo Lucena <plucena at coopergeneral.com>
> wrote:
> >
> > ?I realized I made a typo:
>
>
> switch.ch has a nice bandwidth delay product calculator.
> https://www.switch.ch/network/tools/tcp_throughput/ <
> https://www.switch.ch/network/tools/tcp_throughput/>
>
> Punching in the link spec from the original post, gives pretty much
> exactly what you said Pablo, including that it'd get ~6.999 megabits with a
> default 64k window.
>
> BDP (100 Mbit/sec, 75.0 ms) = 0.94 MByte
> required tcp buffer to reach 100 Mbps with RTT of 75.0 ms >= 915.5 KByte
>
> Theo
>
>
>
>
> ------------------------------
>
> Message: 14
> Date: Fri, 6 Nov 2015 10:35:13 +1100
> From: Greg Foletta <greg at foletta.org>
> To: alvin nanog <nanogml at mail.ddos-mitigator.net>
> Cc: Eric Dugas <edugas at unknowndevice.ca>, nanog at nanog.org
> Subject: Re: Long-haul 100Mbps EPL circuit throughput issue
> Message-ID:
>         <CAN5PdK=sUjKMeRXkqGiEVCv5_Fi3kF3pr_gBYBB=-
> 40S2yjzZg at mail.gmail.com>
> Content-Type: text/plain; charset=UTF-8
>
> Along with recv window/buffer which is needed for your particular
> bandwidth/delay product, it appears you're also seeing TCP moving from
> slow-start to a congestion avoidance mechanism (Reno, Tahoe, CUBIC etc).
>
> Greg Foletta
> greg at foletta.org
>
>
> On 6 November 2015 at 10:19, alvin nanog <nanogml at mail.ddos-mitigator.net>
> wrote:
>
> >
> > hi eric
> >
> > On 11/05/15 at 04:48pm, Eric Dugas wrote:
> > ...
> > > Linux test machine in customer's VRF <-> SRX100 <-> Carrier CPE (Cisco
> > > 2960G) <-> Carrier's MPLS network <-> NNI - MX80 <-> Our MPLS network
> <->
> > > Terminating edge - MX80 <-> Distribution switch - EX3300 <-> Linux test
> > > machine in customer's VRF
> > >
> > > We can full the link in UDP traffic with iperf but with TCP, we can
> reach
> > > 80-90% and then the traffic drops to 50% and slowly increase up to 90%.
> >
> > if i was involved with these tests, i'd start looking for "not enough tcp
> > send
> > and tcp receive buffers"
> >
> > for flooding at 100Mbit/s, you'd need about 12MB buffers ...
> >
> > udp does NOT care too much about dropped data due to the buffers,
> > but tcp cares about "not enough buffers" .. somebody resend packet#
> > 1357902456 :-)
> >
> > at least double or triple the buffers needed to compensate for all kinds
> of
> > network whackyness:
> > data in transit, misconfigured hardware-in-the-path, misconfigured
> iperfs,
> > misconfigured kernels, interrupt handing, etc, etc
> >
> > - how many "iperf flows" are you also running ??
> >         - running dozen's or 100's of them does affect thruput too
> >
> > - does the same thing happen with socat ??
> >
> > - if iperf and socat agree with network thruput, it's the hw somewhere
> >
> > - slowly increasing thruput doesn't make sense to me ... it sounds like
> > something is cacheing some of the data
> >
> > magic pixie dust
> > alvin
> >
> > > Any one have dealt with this kind of problem in the past? We've tested
> by
> > > forcing ports to 100-FD at both ends, policing the circuit on our side,
> > > called the carrier and escalated to L2/L3 support. They tried to also
> > > police the circuit but as far as I know, they didn't modify anything
> > else.
> > > I've told our support to make them look for underrun errors on their
> > Cisco
> > > switch and they can see some. They're pretty much in the same boat as
> us
> > > and they're not sure where to look at.
> > >
> >
>
>
> ------------------------------
>
> Message: 15
> Date: Thu, 5 Nov 2015 23:40:22 -0500
> From: William Herrin <bill at herrin.us>
> To: Pablo Lucena <plucena at coopergeneral.com>
> Cc: "NANOG Operators' Group" <nanog at nanog.org>
> Subject: Re: Long-haul 100Mbps EPL circuit throughput issue
> Message-ID:
>         <
> CAP-guGUycg9mcVKYxB_fXXVCLhNpA6unkCmoiODrFmn5i7aDOw at mail.gmail.com>
> Content-Type: text/plain; charset=UTF-8
>
> On Thu, Nov 5, 2015 at 9:17 PM, Pablo Lucena <plucena at coopergeneral.com>
> wrote:
> > With default window size of 64KB, and a delay of 75 msec, you should only
> > get around 7Mbps of throughput with TCP.
>
> Hi Pablo,
>
> Modern TCPs support and typically use window scaling (RFC 1323). You
> may not notice it in packet dumps because the window scaling option is
> negotiated once for the connection, not repeated in every packet.
>
> Regards,
> Bill Herrin
>
>
> --
> William Herrin ................ herrin at dirtside.com  bill at herrin.us
> Owner, Dirtside Systems ......... Web: <http://www.dirtside.com/>
>
>
> ------------------------------
>
> Message: 16
> Date: Fri, 6 Nov 2015 00:17:26 -0500
> From: Pablo Lucena <plucena at coopergeneral.com>
> To: William Herrin <bill at herrin.us>
> Cc: "NANOG Operators' Group" <nanog at nanog.org>
> Subject: Re: Long-haul 100Mbps EPL circuit throughput issue
> Message-ID:
>         <CAH+2GVLNx8-rxHe9_cHaAv2oVosf=45bh=
> cP6NBesD90Z3EstQ at mail.gmail.com>
> Content-Type: text/plain; charset=UTF-8
>
> >
> >
> > Modern TCPs support and typically use window scaling (RFC 1323). You
> > may not notice it in packet dumps because the window scaling option is
> > negotiated once for the connection, not repeated in every packet.
> >
> >
> Absolutely. Most host OS should support this by now. Some test utilities
> however, like iperf (at least the versions I've used) default to a 16 bit
> window size though.
>
> The goal of my response was to allude to the fact that TCP relies on
> windowing unlike UDP, thus explaining the discrepancies.
>
> This is a good article outlining these details:
>
> https://www.edge-cloud.net/2013/06/measuring-network-throughput/
>
>
> ------------------------------
>
> Message: 17
> Date: Fri, 6 Nov 2015 05:12:29 +0000 (UTC)
> From: Brandon Wade <brandonwade at yahoo.com>
> To: "nanog at nanog.org" <nanog at nanog.org>
> Subject: Re: Internap route optimization
> Message-ID:
>         <671102648.693176.1446786749702.JavaMail.yahoo at mail.yahoo.com>
> Content-Type: text/plain; charset=UTF-8
>
> > Does anyone know or have any experience with Internap's route
> > optimization? Is it any good?
> >
> > I've heard of competing solutions as well, such as the one
> > provided by Noction.
> >
> > Thanks for your input,
> > Paras
>
> We currently utilize the Border6 solution on our network and are very
> happy with it. It optimizes our outbound traffic by injecting more
> specifics into our border router to influence the outbound traffic. To
> assure we don't re-advertise these more specifics to our downstream/peers
> we tag those routes with a community.
>
> The Border6 solution optimizes traffic based upon performance and also
> allows us to keep our traffic levels within our commits. It has saved us on
> bursting costs, as well as technical support costs since it has virtually
> eliminated packet loss complaints.
>
> As far as price, the Border6 solution is by far way more cost effective
> verses the quotes we have received from Internap and Noction. The technical
> support staff have always gone above and beyond to tweak the solution as
> needed as well. Overall we are very impressed and would recommend the
> Border6 solution. It more than pays for itself with customer satisfaction
> and not needing to staff someone around the clock to manually route around
> packet loss, blackholed traffic, etc.
>
> In fact a few weeks ago our appliance died (for reasons not related to the
> Border6 solution) and it took a few days for us to provide a replacement
> box and we did notice complaints come in for packet loss, etc. A pretty
> good indication that it is doing it's job as expected.
>
> Feel free to contact me off list if you have any questions about the
> Border6 solution.
>
> Brandon Wade - AS53767
> http://as53767.net
>
>
>
>
> ------------------------------
>
> Message: 18
> Date: Fri, 6 Nov 2015 08:59:33 +0100
> From: Seth Mos <seth.mos at dds.nl>
> To: NANOG list <nanog at nanog.org>
> Subject: Youtube CDN unreachable over IPv6
> Message-ID: <563C5DE5.60505 at dds.nl>
> Content-Type: text/plain; charset=utf-8
>
> Dear Google,
>
> It appears that one of the Youtube CDN's (in Europe, NL) is not
> reachable over IPv6 from AS 20844. Can someone get back to us on this,
> the company can't access any of the videos currently, although the
> mainpage loads fine (over IPv6).
>
> Kind regards,
>
> Seth
>
> telnet r6---sn-5hne6n76.googlevideo.com 443
> Trying 2a00:1450:401c:4::b...
> telnet: connect to address 2a00:1450:401c:4::b: Connection timed out
> Trying 74.125.100.203...
> Connected to r6.sn-5hne6n76.googlevideo.com (74.125.100.203).
> Escape character is '^]'.
> Connection closed by foreign host.
>
> telnet www.youtube.com 443
> Trying 2a00:1450:4013:c01::5d...
> Connected to youtube-ui.l.google.com (2a00:1450:4013:c01::5d).
> Escape character is '^]'.
> Connection closed by foreign host.
>
>
> End of NANOG Digest, Vol 94, Issue 4
> ************************************
>



More information about the NANOG mailing list