Is there anything to be done with the packet loss?

[sfbmod]

I'm getting <200ms pings to the Seatribe server. But 99% packet loss at the Hetzner host seems insane. I believe this may be the source of the extreme lag and rubber banding we're seeing. Is this just a host issue hopefully?

server ping 4-20-12.JPG

[sfbmod]

Or are those just redundant firewalls that aren't invisible to a public ping like they should be?

[borka]

@sfbmod

can you do a trace with this online site pls http://www.t1shopper.com/tools/traceroute/ (i can't atm cuz my IP gets recognized as a german one so service is refused...) - it should show "multiple routers" at Hetzner ...

btw i got around 175 ms ... 2 days ago when it worked with my IP at t1...

with ping plotter i get packet losses too... 50% in 9 hops...

[sfbmod]

Ok, but that webtool doesn't show packet loss. Things look normal from just a standard ping test:

traceroute to 178.63.100.209 (178.63.100.209), 20 hops max, 40 byte packets
1 208.64.252.229.uscolo.com (208.64.252.229) 0.274 ms 0.258 ms 0.298 ms
2 208.64.248.17.uscolo.com (208.64.248.17) 0.429 ms 0.460 ms 0.508 ms
3 tg1-7--100.br01.lsan.acedc.net (69.27.173.9) 0.370 ms 0.374 ms 0.398 ms
4 tg1-1.ar01.prov.acedc.net (199.58.196.37) 17.302 ms 17.321 ms 17.350 ms
5 67.137.124.85 (67.137.124.85) 19.565 ms 19.557 ms 18.546 ms
6 tg9-2.cr02.slkcutxd.integra.net (209.63.98.209) 86.449 ms 207.340 ms 194.915 ms
7 tg0-3-0-13.sc02.slkcutxd.integra.net (209.63.82.74) 25.102 ms 25.106 ms 25.145 ms
8 be3.sc01.lsancarc.integra.net (209.63.82.94) 19.758 ms 19.779 ms 19.766 ms
9 tg1-1.br01.lsancarc.integra.net (209.63.113.58) 18.501 ms 18.526 ms 18.541 ms
10 init7.net.any2ix.coresite.com (206.223.143.85) 27.481 ms 27.550 ms 27.563 ms
11 r1nyc1.core.init7.net (82.197.168.70) 91.568 ms 92.068 ms 91.657 ms
12 r1lon1.core.init7.net (77.109.140.193) 160.012 ms 160.009 ms 160.009 ms
13 r1fra1.core.init7.net (77.109.128.38) 171.855 ms 171.906 ms 171.874 ms
14 gw-hetzner.init7.net (82.197.166.86) 171.608 ms 171.710 ms 171.679 ms
15 hos-bb1.juniper1.rz12.hetzner.de (213.239.240.251) 179.887 ms hos-bb1.juniper2.rz12.hetzner.de (213.239.240.252) 176.474 ms 177.078 ms
16 hos-tr3.ex3k15.rz12.hetzner.de (213.239.228.208) 177.500 ms 179.542 ms 177.248 ms
17 moltke.seatribe.se (178.63.100.209) 176.560 ms 176.561 ms 176.590 ms

I suspect whatever these extra routers or firewalls are doing may be passing packets around too much causing all kinds of conflicts and collisions. Could they be mirroring ports and just not returning ping packets?

If it's not the host, then is it safe to presume the database is just over tasked? I'm only able to test from a single location so can't verify lag spike variances easily.

[borka]

FYI - have a look here:
http://forums.comcast.com/t5/Connectivi ... -p/1192695

second post by 33tango - the trace looks familiar...

too bad i can't find any informations about Hetzner "internal" routing yet...

hos-bb1.juniper1.rz12.hetzner.de and hos-bb1.juniper2.rz12.hetzner.de

are Juniper EX Core Switches http://www.juniper.net/

[sfbmod]

Thanks for looking into this Borka. I have a friend in Washington State that has the same exact problems typically at the same times of day as well, who also uses Comcast ISP as I and the others on that forum do. They just raised my 'bundle' package another $20 this month, so I'm seriously thinking of changing ISPs. If I do I'll let you know how if it changes anything.

Thanks again.

[sfbmod]

This review from last February is detailing the same problem, but saying it's due to Hetzner:

http://www.webhostingtalk.com/archive/i ... 30371.html

"Connections to the U.S. sometimes suffer packet loss, significant latency, and throughput issues (maximum upstream sometimes capped at ~500 KB/s even to Tier 1 dedicated servers in central US). This is more noticeable the more west you go in the U.S. and is to the point that this is not suitable for gaming if you live outside of the Eastern US time zone.
Their DC Backbone suffers from fairly high utilization and is continually upgraded and then maxed out again, due to the high volume of server sales and generous bandwidth offerings of Hetzner."

[borka]

- snip -
Their DC Backbone suffers from fairly high utilization and is continually upgraded and then maxed out again, due to the high volume of server sales and generous bandwidth offerings of Hetzner."

Where DC is Datacenter (Rechenzentrum - rz12 Falkenstein/Germany for HnH) and not some backbone in USA Capital