Slow when others download?

swendorf1 · ‎07-29-2004

Hello,

We have a very simple network setup. Cisco 2611 with a serial T1. We use only one of the ethernet ports for our internal network.Only have about 15 total machines from a switch that is connected to the ethernet interface of the 2611. OUR PROBLEM: When a single user within our site downloads a file it dramatically slows down all other users in the office and users accessing our Citrix Server from the outside. When I traceroute or ping or serial interface as a download is in progress, the latency is very high (around 400ms). As soon as the download stops, the piing and traceroute times drop back to about 7ms. We used to use a different carrier months ago and never seemed to have this problem. I have checked the cpu and ram usage within the 2611 while a download is on progress they appear to be normal. I have contaceted qwest (our carrier) and they claim there iis no packet loss and all links to us are functioning normally. SO I am just a little stuck here. I''m not sure if this is normal because with our last carrier we did not have this problem running the same speed. I confirmed with our last carrier that they were not running any type of prioritization software also. Can anybody throw some ideas my way that I could chjeck on. I would really appreciate it.

david-schroeder_2 · ‎07-29-2004

Hi,

Have you checked the T1 utilization when the download is occuring? Is it running 100% busy?

Dave

swendorf1 · ‎07-29-2004

Hi,

Yes, Qwest tells me that we our utilization is very high. But why were we fine with the other carrier?

david-schroeder_2 · ‎07-29-2004

Do you have access to the router that the Qwest T1 is attached to? If so, then get into that router and display the Serial interface of that connection. You can then see the 5 minute traffic averages on that interface to be certain if you are using 100% of the pipe during the download. 100% utilized would be 1544kb per sec one way. You need to see that statistic to ensure you are actually getting full T1 throughput.

Dave

Richard Burts · ‎07-29-2004

Were there changes in the router configuration when you switched carriers (I would assume there were) and if so what changed?

It might be beneficial if you posted the configuration of the router.

HTH

Rick

HTH

Rick

swendorf1 · ‎07-29-2004

I just talked to the qwest Tech. He told me that they have verified the usage at THIER router and it is okay...Here is a copy of my config

! Last configuration change at 11:58:48 mst Thu Jul 29 2004

!

version 12.0

service timestamps debug uptime

service timestamps log uptime

service password-encryption

service udp-small-servers

service tcp-small-servers

!

hostname CISCO2600

!

boot system flash c2600-is-mz

enable secret xxx

enable password xxx

!

clock timezone mst -6

ip subnet-zero

ip name-server 205.171.3.65

!

interface Loopback0

no ip address

no ip directed-broadcast

!

interface Ethernet0/0

ip address 10.1.1.254 255.255.255.0

ip access-group 100 in

no ip directed-broadcast

ip nat inside

full-duplex

!

interface Serial0/0

description qwest ds1it-8885442 acct65800901 8778866515

ip address 208.x.x.x.x.255.252

no ip directed-broadcast

ip nat outside

no ip mroute-cache

!

interface Ethernet0/1

ip address 192.168.1.100 255.255.255.0

no ip directed-broadcast

ip nat inside

full-duplex

!

router rip

version 2

redistribute static

network 10.0.0.0

default-metric 1

no auto-summary

!

ip nat pool netpool 208.x.x.x.46.86.154 netmask 255.255.255.252

ip nat inside source list 2 pool netpool overload

ip nat inside source static 10.1.1.7 208.46.86.158

ip classless

ip route 0.0.0.0 0.0.0.0 208.46.78.53

no ip http server

ip http access-class 1

!

logging history size 100

logging history emergencies

logging trap debugging

logging facility syslog

logging source-interface Loopback0

logging 10.1.1.3

access-list 2 permit 10.1.1.0 0.0.0.255

access-list 2 permit 192.168.1.0 0.0.0.255

access-list 100 deny tcp host 10.1.1.7 any eq www

access-list 100 deny tcp host 10.1.1.8 any eq www

access-list 100 permit tcp any any

access-list 100 permit ip any any

banner motd how

!

line con 0

transport input none

speed 115200

line aux 0

line vty 0 4

password xxx

login

!

ntp clock-period 17208690

ntp peer 192.43.244.18

no scheduler allocate

end

david-schroeder_2 · ‎07-29-2004

Can you send me the output of the show int s0/0 command when the download is occuring?

Dave

swendorf1 · ‎07-29-2004

David,

I sent what you needed to your email address. Thank you for your help.

david-schroeder_2 · ‎07-29-2004

Thanks for the output. Check the response time to the device in quesiton while the download is occuring and compare it to the response time to the other side of the T1.

swendorf1 · ‎07-29-2004

The response time is good inside the LAN to the device. Less than 1ms. The response time to the device from outside the router is really high, 500ms. It seems the slow down occurs when you travel a path that runs through a couple of Qwest Routers to our Routers serial port.

david-schroeder_2 · ‎07-29-2004

So if you measure response time from your router to the Qwest router directly downstream from your router via the T1, the response time is good? Then if you measure to the device in question which is furthur downstream from the Qwest router directly downstream from yours you see 400-500ms? I would expect as you tested, no response time issue on your local LAN devices.

If there is more links involved downstream from the Qwest router, then those links are most likely the culprit of your poor response times.

Dave

swendorf1 · ‎07-29-2004

Dave here is a tracert my brother did for me from outside our network using hisss machine as I was downloading something from inside our network,

1 <1 ms <1 ms <1 ms 192.168.0.1

2 40 ms 40 ms 40 ms 63-227-1-254.dnvr.qwest.net [63.227.1.254]

3 47 ms 39 ms 40 ms dnvr-agw1.inet.qwest.net [207.225.112.61]

4 48 ms 40 ms 40 ms thn-core-02.inet.qwest.net [206.196.128.247]

5 49 ms 41 ms 40 ms dia-core-01.inet.qwest.net [205.171.8.81]

6 48 ms 40 ms 40 ms dvr-edge-10.inet.qwest.net [205.171.10.138]

7 373 ms 369 ms 369 ms 208.46.78.54

8 365 ms 361 ms 359 ms citrixserver01.oilexchangevdr.com [208.46.86.158]

Now, so you can get a clear idea...

If I start a download from my machine internally from outside the router and then ping the Qwest Router that is one hop away at 205.171.10.138 the ping time is really good.

But if I ping the next hop after that 205.171.8.81 then immediately the response time is lousy like 400ms.

How can the response time be good from my router to the first router hop then jump so high to the second hop and Qwest tells me that it is simply a saturation problem with my line?

david-schroeder_2 · ‎07-30-2004

Shannon,

Traceroute from your Citrix Server to the IP address of hop 2 or three in your traceroute from the outside and post the results. Based upon the traceroute from the outside in to your Citrix server it would appear the T1 may be the issue. But, since you did extended pings from the inside across the T1 to the Qwest router and they ran well while the download was going, it would appear the T1 would not be the issue. I have had instances where the telco had issues with a circuit on data traveling one way but no the other. The traceroute info from the inside out may depict that.

Dave

vcjones · ‎07-29-2004

Sounds like a classic case of queueing delay. Most likely, your previous service provider was using "fair queueing" or the equivalent while Qwest is using "FIFO" (first in, first out). Talk to Qwest to see if you can get them to change the queuing strategy on your access line. There is nothing you can do from your end if the backup is at their end of the line.

Look up "weighted fair queueing" here on CCO for an outline of the problem and what you can/can't do about it.

Good luck and have fun blackmailing your service rep :-)

Vincent C Jones

www.networkingunlimited.com