Re: [conduit] Large CONDUIT latencies to UW-Madison idd.aos.wisc.edu starting the last day or two.

  • To: Carissa Klemmer - NOAA Federal <carissa.l.klemmer@xxxxxxxx>
  • Subject: Re: [conduit] Large CONDUIT latencies to UW-Madison idd.aos.wisc.edu starting the last day or two.
  • From: Bob Lipschutz - NOAA Affiliate <robert.c.lipschutz@xxxxxxxx>
  • Date: Mon, 22 Feb 2016 17:45:27 +0000
Carissa,

As you know, we've been reporting slow-to-very-slow ftp downloads from NCEP
to NOAA/ESRL/GSD in Boulder for some time, including occasional rates as
low as Patrick's .4 MB/s. At the moment, my curl download of a GFS 0.25 deg
file to /dev/null is running about 2-4 MB/s (~1-2 minutes to download).
That's a fraction of what we might expect over the NOAA NWAVE connection.
And, given that our route is independent of Internet2 and external ISP
paths traversed by UCAR and others, this would seem to suggest an issue
within the NCEP FTP/HTTP dissemination environment...

    Bob Lipschutz
    NOAA/ESRL/Global Systems Division
    IT Services/Data Services Group
    303-497-6636

p.s. Here's our traceroute:

traceroute ftp.ncep.noaa.gov
traceroute to ftp.ncep.noaa.gov (140.90.101.61), 30 hops max, 40 byte
packets
 1  137.75.129.1 (137.75.129.1)  0.274 ms  0.288 ms  0.307 ms
 2  140.172.2.137 (140.172.2.137)  0.169 ms  0.153 ms  0.147 ms
 3  2001-mlx8-esrl-bb.boulder.noaa.gov (140.172.253.254)  0.310 ms  0.336
ms  0.376 ms
 4  radio-rtr-xe-4-2-0-0.boulder.noaa.gov (140.172.2.17)  0.400 ms  0.439
ms  0.435 ms
 5  dsrc-rtr-xe-5-2-1-0.boulder.noaa.gov (140.172.2.26)  0.325 ms  0.365 ms
 0.391 ms
 6  rtr.boul-xe-1-2-0.boulder.noaa.gov (140.172.3.206)  0.366 ms  0.418 ms
 0.327 ms
 7  tge-0-0-0-3.2.rtr2.denv.nwave.noaa.gov (140.172.88.16)  43.899 ms
 43.902 ms  43.887 ms
 8  tge-0-0-0-0.rtr.denv.nwave.noaa.gov (140.172.70.18)  44.035 ms  44.016
ms  44.022 ms
 9  140.172.70.48 (140.172.70.48)  43.164 ms  43.182 ms  43.396 ms
10  140.208.63.29 (140.208.63.29)  43.398 ms  43.381 ms  43.385 ms
11  140.208.63.30 (140.208.63.30)  85.863 ms  85.899 ms  85.798 ms
12  140.90.111.36 (140.90.111.36)  91.453 ms  64.929 ms  64.878 ms
13  140.90.76.69 (140.90.76.69)  51.341 ms  51.680 ms  51.270 ms


On Mon, Feb 22, 2016 at 2:50 PM, Carissa Klemmer - NOAA Federal <
carissa.l.klemmer@xxxxxxxx> wrote:

> Patrick,
>
> So from the looks of your screen grab that is 7 minutes to download the
> 0.25 GFS from our FTP server. Do you have ideas of how slow that is
> compared to normal from Colorado? I will be passing this information on to
> our IT folks.
>
> Carissa Klemmer
> NCEP Central Operations
> Dataflow Team Lead
> 301-683-3835
>
> On Mon, Feb 22, 2016 at 8:52 AM, Patrick L. Francis <wxprofessor@xxxxxxxxx
> > wrote:
>
>>
>>
>> All,
>>
>> Something bigger than just CONDUIT users occurred on the 19th. We had
>> numerous users of other servers complain of similar drops in transfer
>> rates. A few of them were also able to move downloads to a new location
>> where speeds were normal. We believe that this issue was outside of NCEP
>> due to that reasoning. Having said that though, is anyone still seeing
>> abnormal rates?
>>
>>
>>
>>
>>
>> Hi Carissa J
>>
>>
>>
>> From our Amazon location in VA, everything is fine as it was on Fridayâ
>> From our colo facility, packet loss is still shown via 140.90.111.36, which
>> is the hop from gigapop to ncep.. here is an MTR showing the packet loss:
>>
>>
>>
>> http://drmalachi.org/files/ncep/he-ncep.2016.02.22.jpg
>>
>>
>>
>> Notice that since our colo is directly on the Hurricane Electric
>> backbone, there are only a few hops to get to the ncep server, and pings
>> only jump once they reach gigapop.
>>
>>
>>
>> Here is a screen cap of current download speeds for 0.25deg gfs from the
>> ncep server to our colo:
>>
>>
>>
>> http://drmalachi.org/files/ncep/he-ncep.wget.2016.02.22.jpg
>>
>>
>>
>> In order to ensure proper delivery, we have setup a push / pull from our
>> amazon box to the colo, but most people probably wonât have that
>> flexibility.. The issue appears to potentially be related to
>> packetfiltering and / or redirection / funneling via certain routes, but
>> thatâs just a guess J
>>
>>
>>
>> Happy Monday J
>>
>>
>>
>> Cheers,
>>
>>
>>
>> --patrick
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> -------------------------------------------------------
>>
>> Patrick L. Francis
>>
>> Vice President of Research & Development
>>
>>
>>
>> Aeris Weather
>>
>>
>>
>> http://aerisweather.com/
>>
>> http://modelweather.com/
>>
>>
>>
>> http://facebook.com/wxprofessor/
>>
>>
>>
>> --------------------------------------------------------
>>
>
>
> _______________________________________________
> conduit mailing list
> conduit@xxxxxxxxxxxxxxxx
> For list information or to unsubscribe, visit:
> http://www.unidata.ucar.edu/mailing_lists/
>