Topic: Excessively long ping times

[edb]

Well I discovered something interesting with the Colasoft graphical ping tool ...
I went to http://www.speakeasy.net/speedtest/ and ran a test from any of the servers listed there and I watched the ping tool results as I conducted the test. I have repeated this several times.
What I saw happening was that on the download portion of the speed test the ping times remained stable but as soon as it began the upload portion of the test and until it was over I witnessed complete packet loss (just a big blank space of time on the screen). Once the tests were completed the charted ping times returned to normal again.

This may explain the totally eradic problem with the long ping times.
I have seven site-to-site VPNs running as well as an ecommerce/web/mail server and I'm starting to think that it is the lack of upload capacity that is causing my issue. We have a monthly flyer which can be downloaded by our online customers and it is usually around 10-12 MB in size. Right now we don't have all that much traffic to our online store but It may be the combination of bandwidth starvation when it comes to anything upload related.

comments?

edb

[mercyh]

Interesting....

I replicated your testing on my connection. (2.8mbps symmetrical DSL) I normally get about 1.4 actual on speed tests with upload running just a little faster then download.

The ping times remained flat during the download side of the test but doubled on the upload side. Ping to www.kansas.com was around 45Ms download and 90Ms upload. The speed test times showed 1388kbps download and 1620kbps upload. I don't understand how traffic is routed but why does upload apparently take precedence over download traffic? (upload seems to saturate the bandwidth but download does not.) I am not running any traffic shaping on my network and I presume both upload and download is happening on the same protocol (http).

Ping to the ISP gateway. (The first router on the WAN side of my modem) Was 5ms prior to the test, 20ms during the download portion of the test and 79ms during the upload part of the test.

Ping to my LAN gateway. (My router) remained 1ms throughout the test. 

[edb]

That is Interesting mercyh

I'm surprised by your upload speed being as high as it is becuase usually it's throtled.
I wonder if anyone with a T1 connection would get the same results being a none shared connection?
My assumption is that it would be an unchanged up & down ping time but I don't have a T1 yet to test.

edb

[mercyh]

My ISP is a small town (population 3,000) phone Co. that writes their own rules. They are the only broadband isp here other then radio or satellite. They truly are selling an unthrottled upload/download connection, however the bandwidth is still shared among all customers. I think they are currently serving from around 12mbs of paired ds3/t1s. Our 1:00pm speadtests bog down to around 450kbs.

[edb]

Ok, that makes sense .... I'm thinking of getting a bonded T1 connection to give me 3.08Mbps both up and down but I'm just trying to confirm that it would actually make a big difference in the end and that it's not an internal networking issue on my end.
Don't want to just throw money at it unless the T1 would actually help becuase as far as download speeds they are comparable to cable but much better on the upload side due to cables throtling.
Since I'm starting to believe that my issue is upload related a T1 may be my fix.

edb

[raem]

edb

Perhaps this would help
http://wiki.contribs.org/Wondershaper

[mercyh]

Would wondershaper on SME help if the SME is not the gateway router?

[raem]

Would wondershaper on SME help if the SME is not the gateway router?

Well in the very first post edb says "I have my SME server in gateway mode on a Sonicwall 2040 - DMZ port with a public IP and a private IP on the internal nic."

If another router is the gateway, then that may need outbound traffic shaping.


Also see this post re wondershaper & ping times, it may be a clue to resolving your problems.
http://forums.contribs.org/index.php?topic=29546.30

[edb]

Thank for that tip Ray ...

I don't know much about traffic shaping but I'll look into that.
I think the Sonicwall 2040 has it's own bandwith management and QOS stuff but I've not looked at what it can do in that area because I didn't think it would help to any great extent.

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.

edb

[edb]

Here is an example of when things are seemingly running smoothly and I run the http://www.speakeasy.net/speedtest/ speed test. What you see in the graph is the increased ping times when it begins the upload test and until it finishes. Note that the ping times went from 120ms to 2000+ms for that short period of time. This is the kind of thing that happens regularly and kills my Point-Of-Sale VPN connections when it does.
It can happen randomly for a few seconds or several minutes at a time.
Like finding a needle in a haystack as to what the cause is.

[imcintyre]

<~~not an expert but aren't "no load" ping times of 120 ms on an internal network excessive?

I'm on a wireless AP --- switch ---- gateway--modem (+some other stuff). No gigabit hardware anyware and wireless G. I have 1- 2 ms to get to gateway router, to my home network, I have 40 ms avg.

I ran the same speedtest and got Toronto - New York -   3235 kbps down and 3730 kbps up with almost no change internally and maybe doubling of external, see graph. Well I would show you the graph if I could get it into the post. I tried to cut and past bitmap and jpeg but neither would work.

Anyways I tried just pinging internal addresses with the speedtest running and I don't see the ping time to gateway at more than 2 ms for either up or down.

I guess I'm saying that something else is wrong inside the network, not at the gateway/firewall/modem level.

[mercyh]

I guess I'm saying that something else is wrong inside the network, not at the gateway/firewall/modem level.

I would agree except that I can't tell from the IP's if these are all targets that are across the vpn tunnels. If one of these targets is on the local network there is something strange inside.

[edb]

<~~not an expert but aren't "no load" ping times of 120 ms on an internal network excessive?

These internal IPs are my site-to-site VPN tunnels through the Sonicwall to other store locations.

[mercyh]

add the internal IP of the sonicwall itself to your ping list I would assume this is your LAN internet gateway address. Also add the Sonicwall's gateway IP address on the Wan side. (This should be the first router past the sonicwall.) If the problem is on the cable side the internal IP address will not increase it's ping time but the external one will. If the problem is on the LAN side both pings will increase.

[edb]

I think my issue may have a lot to do with email attachments.
We are a sales oriented company so the sales reps are doing quotes and sending customers pdf attachments that can sometimes be several MB in size.
Also, our sales reps receive pdf files from our suppliers regularly as well through email.
Here is the graph when I send a 1MB pdf email attachment to myself.

Anyone else seen anything like this?