Ok here we go!
I have a server on the regular network, and I have a server on the Cogent network.
One is windows, one is RHEL4.....Windows is on the regular network, linux server is on the cogent network.
When I download a file off of the linux server to the windows server, I can't get above 250Kbps download speed.
If I download a file off of the windows server to the linux server, I can easily hit 80+Mbps.
I can hit my full 12Mbps download off of my cable connection.
Any ideas??
btw:
4475494PLNT
Full Version: What do you think?
QUOTE (thedude @ Dec 28 2007, 03:47 PM) 
Ok here we go!
I have a server on the regular network, and I have a server on the Cogent network.
One is windows, one is RHEL4.....Windows is on the regular network, linux server is on the cogent network.
When I download a file off of the linux server to the windows server, I can't get above 250Kbps download speed.
If I download a file off of the windows server to the linux server, I can easily hit 80+Mbps.
I can hit my full 12Mbps download off of my cable connection.
Any ideas??
btw:
4475494PLNT
I have a server on the regular network, and I have a server on the Cogent network.
One is windows, one is RHEL4.....Windows is on the regular network, linux server is on the cogent network.
When I download a file off of the linux server to the windows server, I can't get above 250Kbps download speed.
If I download a file off of the windows server to the linux server, I can easily hit 80+Mbps.
I can hit my full 12Mbps download off of my cable connection.
Any ideas??
btw:
4475494PLNT
Hi there!
Could you run a traceroute or two... from each box to the other? This could help identify the problem. I've updated the ticket with this request as well.
Replied, and for the heck of it I'll post the results here as well.
I don't see anything really in the trace that looks wrong.
One other thing:
If someone here wants to download something off of my server, and let me know what download speed you get, that would be good probably.
Download from one of your TP/EV1 servers and post the results here.
http://webmesto.net
Don't laugh at my site...its just a cover page I did at 1:00 A.M to hide the directory...lol
Feel free to grab any of the ISO's and what not I have linked on there.
I don't see anything really in the trace that looks wrong.
QUOTE
Linux / Cogent server to Windows server on regular network:
[root@srv1 csf]# traceroute 74.52.164.x
traceroute to 74.52.164. (74.52.164.x), 30 hops max, 38 byte packets
1 d9.d.5546.static.theplanet.com (70.85.13.217) 0.874 ms 0.832 ms 0.850 ms
2 2.7c.5d45.static.theplanet.com (69.93.124.2) 0.401 ms 0.447 ms 0.503 ms
3 vl1.car02.dllstx6.theplanet.com (12.96.160.23) 4.816 ms 3.037 ms 3.128 ms
4 82.a4.344a.static.theplanet.com (74.52.164.x) 0.274 ms 0.623 ms 0.356 ms
Windows server to Linux server on cogent network:
Microsoft Windows [Version 5.2.3790]
© Copyright 1985-2003 Microsoft Corp.
C:\Documents and Settings\Administrator>tracert 70.85.13.x
Tracing route to da.d.5546.static.theplanet.com [70.85.13.x]
over a maximum of 30 hops:
1 <1 ms <1 ms 1 ms 81.a4.344a.static.theplanet.com [74.52.164.129]
2 <1 ms <1 ms <1 ms vl2.dsr02.dllstx2.theplanet.com [12.96.160.42]
3 <1 ms <1 ms <1 ms 5.7c.5d45.static.theplanet.com [69.93.124.5]
4 <1 ms <1 ms <1 ms da.d.5546.static.theplanet.com [70.85.13.x]
Trace complete.
C:\Documents and Settings\Administrator>
[root@srv1 csf]# traceroute 74.52.164.x
traceroute to 74.52.164. (74.52.164.x), 30 hops max, 38 byte packets
1 d9.d.5546.static.theplanet.com (70.85.13.217) 0.874 ms 0.832 ms 0.850 ms
2 2.7c.5d45.static.theplanet.com (69.93.124.2) 0.401 ms 0.447 ms 0.503 ms
3 vl1.car02.dllstx6.theplanet.com (12.96.160.23) 4.816 ms 3.037 ms 3.128 ms
4 82.a4.344a.static.theplanet.com (74.52.164.x) 0.274 ms 0.623 ms 0.356 ms
Windows server to Linux server on cogent network:
Microsoft Windows [Version 5.2.3790]
© Copyright 1985-2003 Microsoft Corp.
C:\Documents and Settings\Administrator>tracert 70.85.13.x
Tracing route to da.d.5546.static.theplanet.com [70.85.13.x]
over a maximum of 30 hops:
1 <1 ms <1 ms 1 ms 81.a4.344a.static.theplanet.com [74.52.164.129]
2 <1 ms <1 ms <1 ms vl2.dsr02.dllstx2.theplanet.com [12.96.160.42]
3 <1 ms <1 ms <1 ms 5.7c.5d45.static.theplanet.com [69.93.124.5]
4 <1 ms <1 ms <1 ms da.d.5546.static.theplanet.com [70.85.13.x]
Trace complete.
C:\Documents and Settings\Administrator>
One other thing:
If someone here wants to download something off of my server, and let me know what download speed you get, that would be good probably.
Download from one of your TP/EV1 servers and post the results here.
http://webmesto.net
Don't laugh at my site...its just a cover page I did at 1:00 A.M to hide the directory...lol
Feel free to grab any of the ISO's and what not I have linked on there.
~ $ wget http://www.webmesto.net/download/SUSE-10.0...iso--19:41:15-- http://www.webmesto.net/download/SUSE-10.0...i386-GM-CD1.iso
=> `SUSE-10.0-CD-OSS-i386-GM-CD1.iso'
Resolving www.webmesto.net... 70.85.13.222
Connecting to www.webmesto.net|70.85.13.222|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 629,899,264 (601M) [application/octet-stream]
3% [> ] 24,304,524 406.00K/s ETA 23:43
Grabbing your ticket.
=> `SUSE-10.0-CD-OSS-i386-GM-CD1.iso'
Resolving www.webmesto.net... 70.85.13.222
Connecting to www.webmesto.net|70.85.13.222|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 629,899,264 (601M) [application/octet-stream]
3% [> ] 24,304,524 406.00K/s ETA 23:43
Grabbing your ticket.
Thanks tomy <3
I pushed the ticket over to our networking group to take a look at the policy-map definitions on the switch to see if they're setup as intended. This would explain the one-way slow down. When I pulled the service-policy definition from your port's config, the transfer pegged the link speed out.
100%[==========================================>] 629,899,264 10.54M/s ETA 00:00
100%[==========================================>] 629,899,264 10.54M/s ETA 00:00
QUOTE (TP-TDurden @ Dec 28 2007, 08:25 PM)
I pushed the ticket over to our networking group to take a look at the policy-map definitions on the switch to see if they're setup as intended. This would explain the one-way slow down. When I pulled the service-policy definition from your port's config, the transfer pegged the link speed out.
100%[==========================================>] 629,899,264 10.54M/s ETA 00:00
100%[==========================================>] 629,899,264 10.54M/s ETA 00:00
I know its a 20Mbps port, and I've only been able to push 20Mbps out to the internet, but inbound, I've been able to max out my port..lol
And thanks for looking into it Tomy..I appreciate it.
Hi -
The behavior you're seeing is because of how traffic policing works on the switch. We can apply policing policies to your port only on the inbound side (this is a limitation of the hardware). Because we are using policing and not shaping (most vendors don't support shaping these days because of the memory requirements to do it effectively), what you're seeing is the result of the way the policer does it's magic. Here's a link that might help explain things better:
http://www.cisco.com/warp/public/105/policevsshape.html
If you want a more detailed explanation, let me know and when I have the time I'll send you one.
The behavior you're seeing is because of how traffic policing works on the switch. We can apply policing policies to your port only on the inbound side (this is a limitation of the hardware). Because we are using policing and not shaping (most vendors don't support shaping these days because of the memory requirements to do it effectively), what you're seeing is the result of the way the policer does it's magic. Here's a link that might help explain things better:
http://www.cisco.com/warp/public/105/policevsshape.html
If you want a more detailed explanation, let me know and when I have the time I'll send you one.
I tested the Knoppix download on the main page.
On my Dallas server with 100Mbps port, I got an average of 183Kbps.
On my Bellsouth ADSL 3Mbps port, I got an average of 267Kbps.
On my Dallas server with 100Mbps port, I got an average of 183Kbps.
On my Bellsouth ADSL 3Mbps port, I got an average of 267Kbps.
So because of the traffic policing, thats causing traffic destined for the internal network to be throttled even more outbound?
When you get time I would appreciate the detailed explanation as well.
I see that policing acts more like an iptable in effect due to if traffic reaches the maximum allowed rate in the config of the switch then that traffic is dropped, versus traffic being qued and stored in memory in a traffic shaping environment.
However, shouldn't I still be able to send out more than the 250Kbps to my other TP server, as when I've tested this, I've had more than 10Mbps available on that port? Or maybe I'm reading it wrong at 1 in the morning..lol
Since traffic is controlled on the switch/port that the server is connected to, shouldn't all outbound traffic be treated the same?
Thanks
When you get time I would appreciate the detailed explanation as well.
I see that policing acts more like an iptable in effect due to if traffic reaches the maximum allowed rate in the config of the switch then that traffic is dropped, versus traffic being qued and stored in memory in a traffic shaping environment.
However, shouldn't I still be able to send out more than the 250Kbps to my other TP server, as when I've tested this, I've had more than 10Mbps available on that port? Or maybe I'm reading it wrong at 1 in the morning..lol
Since traffic is controlled on the switch/port that the server is connected to, shouldn't all outbound traffic be treated the same?
Thanks
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.