Topic: Excessively long ping times

[imcintyre]

I just did an experiment that may help. I did the speedtest and had 3206 kbps down and 3807 up.

I connected by openvpn to home network, streamed some music and had 622 down 525 up. During speed test, streamed music started to break up but went when speed test was over. Internal pings did not move at all and external pings were only marginally lengthed if at all.

I stopped streaming but left vpn connection open and had 625 down and 603 up. So the VPN connection has a fair amount of "overhead" that I did not expect

I disconnected vpn and have 3800 down and 3600 up.

Perhaps the experts have something re traffic shapping.

Hope that this is helpful.

[mercyh]

Are you onsite with the mailserver or is it across the VPN tunnel for you??

Try your speedtest with the following tool.

http://www.pingplotter.com/freeware.html

download from the link at the bottom of the page.

install and set your trace to google.com or some other such target, set your trace delay to 2.5 seconds, fire up your speed test and you can see which hop is causing your problem.

This is a graphical traceroute utility. I would expect the problem to be at the first hop after the sonicwall if the cable provider is the problem.

[imcintyre]

I sent an 6 meg file to my home network and ran the speed test. At work I have an external mail server. Internal/External pings were not affected. Speed test results for upload were down by approx 1000 kbps.

[edb]

Here is the graph where I sent an email with a 1MB pdf attachment from work to my home account.
Notice the internal side Sonicwall gateway(192.168.10.88) is uneffected but everything else is.

[edb]

Here is the ping plotter result of sending the same email while tracking www.google.ca

[mercyh]

Did you run these on the same machine???

On pingtool you show your gateway as 192.168.10.88 on the pingplotter graph you have the first hop as ISP's router When I run pingplotter I get the first hop as 192.168.123.254> My LAN internet gateway. My next hop is my ISP's router. When you run IPCONFIG /all from the command line of your workstation do you have the correct gateway shown??

If that pingplotter graph is not obfuscated I think you have a DHCP server that is handing out the wrong gateway address. (Or I am not following your topography at all which is highly possible)

[edb]

Your are correct in that 192.168.10.88 is my computers default gateway.

Here is my ifconfig from my workstation

 

Code: [Select]

IP Address. . . . . . . . . . . . : 192.168.10.30
 Subnet Mask . . . . . . . . . . . : 255.255.255.0
 Default Gateway . . . . . . . . . : 192.168.10.88
 DNS Servers . . . . . . . . . . . : 192.168.10.2
If I do a tracert from the DOS command line ... I get the same result with the ISP's router as the first hop.

[raem]

edb

I also had a look at this thread  http://forums.contribs.org/index.php?topic=40019.0 which talks about qmail adding to bandwidth starvation. Any relevance I don't know.

Is sme server the mail server ?

It's very easy to change the ConcurrencyRemote setting, to say 5 (or less) in your case, and see what happens.
qmail wil definitely utilise all available bandwidth, so the ConcurrencyRemote (& ConcurrencyLocal) settings will definitely tame this behaviour.

[mercyh]

Here is the graph where I sent an email with a 1MB pdf attachment from work to my home account.
Notice the internal side Sonicwall gateway(192.168.10.88) is uneffected but everything else is.

I would go to the ISP with this graph. It seems to me that your mail server is not really the issue as your bandwidth saturates on any sizable upload. Even if you keep the mailserver from using all your bandwidth you will still have the problem with other downloads from your site. IMO if 1Meg of upload will do this to you your provider has oversold their available bandwidth.

[edb]

Yes, that's what I figure as well mercyh so I will provide them with this info but I think I will just change to a bonded T1(3Mb) connection through Bell Canada instead if they can't resolve the bandwidth issue for me.

Ray

What exactly does the ConcurrencyRemote and Local setting do and how does it affect bandwidth?
We are in fact running SME as our mail server but our email goes though a Barracuda Spam firewall first and is then passed through to the SME for all incoming mail. I only have smtp port 25 open to the Baracuda and all mail is sent from the SME server via SSMTP port 465 and remote users use SPOP3 (995) to receive their email.

I'm just curious how the Concurrency settings change things.

edb

[raem]

edb

What exactly does the ConcurrencyRemote and Local setting do and how does it affect bandwidth?

It sets the max number of concurrent messages being processed by qmail (for remote & local messages).
Without a limit qmail will try to process all messages as quickly as possible therefore using/saturating all available bandwidth, and create problems (slowness) for web users, file downloads etc, while those messages are being processed.

[mercyh]

EDB,

You need to run a WHOIS on that ISP Router that you at first showed the IP address on. I was working on something else last night and ran into an IP in that Class A subnet. That block is not located in your country as near as I can tell.

If you have interest in taking this discussion further you can PM me.

rh at mercyh dot org

[raem]

edb

I'll remind you of what you said in your very first post.

"If at the time this is happening I reboot the SME server the ping times quickly return to 45ms again but when the server comes back up it is back to the 2000+ms again. Sometimes this last for a couple of minutes and sometimes it lasts for an hour or more."

Isn't this suggestive of a sme setting or conflict, rather than an external issue.

[edb]

edb

I'll remind you of what you said in your very first post.

"If at the time this is happening I reboot the SME server the ping times quickly return to 45ms again but when the server comes back up it is back to the 2000+ms again. Sometimes this last for a couple of minutes and sometimes it lasts for an hour or more."

Isn't this suggestive of a sme setting or conflict, rather than an external issue.

Yes, Ray that is what I had originally thought as well.
However, as the slowdown was happening the other day (and as I was monitoring with ping) I decided to reboot the SME to confirm that the SME may be the issue. To my surprize, with the SME offline the stats didn't change so that is what led me to it just being bandwidth saturation which may also stem from the 7 site-to-site VPNs that I have running through my Sonicwall. I spoke with my cable ISP yesterday and confirmed that I only have 60K upload capacity and no increase to that is forcast for at least six months.
That being said I think it is just a fact that my setup requires much more upload capacity than what I have available.
I did other research in the last couple days which also confirms this and the problems with cable when any upload capacity is required.
Just a note (I have never had any issues with download bandwidth) downloads have always been stable.

I will be moving us over to a bonded T1 which will provide me with dedicate (not shared) 3Mb up & down capacity and this should resolve my issues.

edb

[edb]

You need to run a WHOIS on that ISP Router that you at first showed the IP address on. I was working on something else last night and ran into an IP in that Class A subnet. That block is not located in your country as near as I can tell.

Actually, it is located in Calgary, AB Canada