[v6test] Test download server

Gert Doering gert at space.net
Mon Mar 24 10:38:08 UTC 2008


Hi,

On Sun, Mar 23, 2008 at 07:14:42AM -0500, Kevin Day wrote:
> I'm starting to do some performance/load testing of our download  
> cluster. If anyone wants to see what kind of speeds they're getting  
> from us right now, there are a couple of videos posted at http://testdownload.ipv6experiment.com 
> . Accessible both over v4 and v6, to help with speed comparisons.
>
> If you're getting really bad or really good download speeds off that,  
> let me know off list. I'm at the point where I can start looking for  
> routing/performance problems to get them fixed before it launches.

IPv4 from me to you has 7 hops, IPv6 has 13, which is a bit surprising
(but most likely there is MPLS hop hiding going on inside pccwbtn...):

traceroute to testdownload.ipv6experiment.com (216.14.100.2), 64 hops max, 52 byte packets
 1  Cisco-M-XXI-Vlan540.Space.Net (194.97.144.214)  73.461 ms  196.680 ms  1.155 ms
 2  ge-1-2-0-zcr2.muc.cw.net (62.208.255.161)  0.373 ms  0.400 ms  0.456 ms
 3  so-4-3-0-dcr2.fra.cw.net (195.2.2.81)  7.052 ms  7.048 ms  7.034 ms
 4  xe-11-3-0.xcr1.fra.cw.net (195.2.9.249)  6.962 ms  6.903 ms  6.898 ms
 5  cr02.frf02.pccwbtn.net (80.81.192.50)  103.076 ms  210.049 ms  14.429 ms
 6  your.org.ge2-5.br02.chc01.pccwbtn.net (63.218.5.38)  121.711 ms  120.361 ms  119.921 ms
 7  testdownload.ipv6experiment.com (216.14.100.2)  121.872 ms  119.249 ms  119.116 ms

traceroute6 to testdownload.ipv6experiment.com (2001:4978:8:1:218:8bff:fe52:b5b0) from 2001:608:0:1007::1007, 64 hops max, 12 byte packets
 1  Cisco-M-XXI-Vlan450.Space.Net (2001:608:0:1007::1)  0.535 ms  0.35 ms  0.322 ms
 2  ge-1-2-0-zcr2.muc.cw.net (2001:5001:100:17::1)  0.743 ms  0.54 ms  0.485 ms
 3  so-4-3-0-dcr2.fra.cw.net (2001:5000:0:cd::1)  7.17 ms so-6-3-0-dcr2.fra.cw.net (2001:5000:0:63::1)  7.254 ms so-4-3-0-dcr2.fra.cw.net (2001:5000:0:cd::1)  7.141 ms
 4  xe-11-3-0.xcr1.fra.cw.net (2001:5000:0:10f::1)  7.057 ms  7.132 ms  7.046 ms
 5  xe-11-0-0.xcr1.par.cw.net (2001:5000:0:105::2)  15.73 ms xe-11-2-0.xcr1.par.cw.net (2001:5000:0:10a::2)  15.747 ms xe-11-1-0.xcr1.par.cw.net (2001:5000:0:109::2)  15.691 ms
 6  xe-11-0-0-xcr1.prp.cw.net (2001:5000:0:ff::1)  47.218 ms  15.834 ms  15.832 ms
 7  xe-11-0-0-xcr1.lnd.cw.net (2001:5000:0:fd::2)  27.907 ms  27.74 ms xe-11-1-0-xcr1.lnd.cw.net (2001:5000:0:fe::2)  27.941 ms
 8  ge-11-0-0.xcr1.nyk.cw.net (2001:5000:0:114::2)  104.807 ms  99.265 ms ge-11-1-0.xcr1.nyk.cw.net (2001:5000:0:115::2)  129.86 ms
 9  2001:504:1::a500:4436:1 (2001:504:1::a500:4436:1)  102.787 ms  102.934 ms  102.288 ms
10  xe-2-1-0.cr2.ord1.us.nlayer.net (2001:590::4516:8e06)  172.439 ms  172.585 ms  171.835 ms
11  tge2-1.ar1.iad1.us.nlayer.net (2001:590::451f:1f8a)  171.892 ms  171.846 ms  172.689 ms
12  yourorg.106.g3-13.ar1.ord1.us.nlayer.net (2001:590::451f:6f16)  117.79 ms  117.634 ms  118.624 ms
13  unassigned.v6.your.org (2001:4978:8:1:218:8bff:fe52:b5b0)  117.316 ms  117.21 ms  117.422 ms

... but the overall RTTs are similar, so let's see how HTTP performs
(using the same URL as Iljitsch)...

gert at delta$ wget -4 -O /dev/null http://testdownload.ipv6experiment.com/24-h2 >
...
11:33:28 (122.09 KB/s) - `/dev/null' saved [36228544/36228544]

gert at delta$ wget -6 -O /dev/null http://testdownload.ipv6experiment.com/24-h2 >
...
11:36:32 (262.17 KB/s) - `/dev/null' saved [36228544/36228544]

So - your IPv4 performance is sucky.  IPv6 is cool :-)

My tests have been run from delta.greenie.net, 194.97.144.209, and
2001:608:0:1007::1007.  NetBSD 2.1 on a Sun Ultra 10 - not the fastest
machine in the world, but should be able to saturate 50 Mbit/s or more...

Gert Doering
        -- NetMaster
-- 
Total number of prefixes smaller than registry allocations:  110584

SpaceNet AG                        Vorstand: Sebastian v. Bomhard
Joseph-Dollinger-Bogen 14          Aufsichtsratsvors.: A. Grundner-Culemann
D-80807 Muenchen                   HRB: 136055 (AG Muenchen)
Tel: +49 (89) 32356-444            USt-IdNr.: DE813185279


More information about the V6test mailing list