WAN Link Interuption

[adamgibs7]

Dears,

I have a link of 7 mbps upload and download speed of wireless point to point connection (layer 2) for my branch, when I try to copy 5GB file on that link the link consumption reaches to 100% and the file copying breaks after sometime, 

Q1.

Though the link get fully consumed by 5GB transfer still the copying should not fail ??? please correct me if I m not wrong it should keep on copying rather than breaking, so if it breaking this mean the point to point link is instable ??

Q2

Which free tool I can use to proof to ISP that their link is not stable.

Thanks

[Joseph W. Doherty]

Q1 a stack's TCP might "fail" if your physical handoff is much, much faster than your path's bandwidth. It shouldn't fail, but I've seen some stacks "give up" if they encounter too many errors.

Do smaller transfers work? If you devices support shaping, try that.

[adamgibs7]

Dears,

Do smaller transfers work? If you devices support shaping, try that

yes smaller transfer work 1GB file, I have not done shaping , the link is on switch and the layer 3 interface is on firewall, on switch it is layer 2 trunk.

what bandwidth measurement tool can justify why the transfer is failing ?? IPERF can be helpful here or some other tool as per your experience.

Regards

[Joseph W. Doherty]

"the link is on switch and the layer 3 interface is on firewall, on switch it is layer 2 trunk."

Shaping, then, might be an unsupported option with your equipment.

[adamgibs7]

Dear

what is your suggestion for the above issue.

thanks

[Joseph W. Doherty]

You might try policing the output to your CIR - Cisco switch often support policing.

Try a different (kind of) host, or, if available, try different TCP stack on host, especially for the sender.

Obtain a device that support shaping.

Obtain a traffic management/shaping appliance.

[Richard Burts]

I think we need to know more about this environment. For example if there is a routing protocol running over this link then the 5G copy using 100% of bandwidth could cause routing protocol packets to be dropped and this could withdraw the routes learned over this link, which could break the copy.

HTH

Rick

[Joseph W. Doherty]

Rick, good point, especially if BGP is being used as I understand it doesn't obtain internal PAK priority.

To OP, are you running a dynamic routing protocol? If so, any log of it dropping when this happens?

[adamgibs7]

Dears

As mentioned above it is a layer 2 trunk with no routing protocols, so what I understand is that if a 5 GB copy is initiated on 7 Mbps link then it is not always success to have copied every time, it might stop copying due to full bandwidth or interruption or link flutuation

I want a reason why the copy is not completing a visualize proof for the failure which software can help me the best, Iperf can help me to justify why it is failing ??

Thanks

[Joseph W. Doherty]

Yes, you did mention it's L2, but without knowing the whole topology setup, it's possible a higher level protocol, such as a routing protocol, is still being impacted by congestion on this part of the end-to-end path.

So are you saying, it's the same L2 domain end-to-end?

Understand, this type of problem can sometimes be difficult to identify the root cause. For example, using IPerf might work fine, as it's not your file copy. (Of course, if you can get such a tool to fail the same way, that would be helpful.)

As an example, many years ago I was working on a somewhat similar problem. Updating the host's NIC driver (with a release that was literally only out for a week), resolved the issue. (BTW, the driver hadn't been recommended, but in this case, when dealing with something that needed fixing because it wasn't working correctly, I thought to try the very latest software for all components.)

[Richard Burts]

I understand that it is a layer 2 connection acting as a trunk. But that does not mean that there is no routing protocol. You have not told us much about your environment so we can not be sure. But think about this as a possible situation. Let us think about an environment where there is a client connected on one side which wants to perform a copy from a server on the other side. There may be several vlans on the trunk and one of those vlans is vlan 10. There is a layer 3 device connected on vlan 10 (could be a router, could be a layer 3 switch, might be something else) which provides connectivity for the client. The layer 3 device runs a routing protocol on vlan 10 to learn the subnet where the server is connected on the other side. The client initiates the copy and the copy consumes the bandwidth on the layer 2 link. Routing protocol packets get dropped because of the congestion. When several routing protocol packets are dropped the routing protocol loses the neighbor relationship with its peer on the other side and withdraws the routing entry for the subnet where the server is. This causes the copy to fail.

HTH

Rick

[adamgibs7]

Dears

I apologize If any of the sentences is disrespecting you'll in the post,

I have restricted to only 2 vlans and rest all are removed and it is a completely layer 2, I want to justify the client why it is failing how I can ??

thanks

[Richard Burts]

We still do not know enough to be able to identify the cause of this issue. I have made a suggestion that there might be some layer 3 protocol traffic (such as a routing protocol) that might be impacted by congestion on the circuit. Larry Sullivan makes a very interesting suggestion that the issue might involve a layer 2 control protocol that controls the wireless circuit that is impacted by congestion. What I think we understand is that there is some device at this site doing a copy. This device is connected to a switch which is connected to a firewall. The firewall is connected over a point to point wireless to some device at the other site which is connected to something (which might be a router or switch) which might be connected to something (might be another switch or might be the other device doing the copy). Any one of those points might be involved in this issue. I would suggest that as the next step in this troubleshooting would be to post the config of the switch and the firewall at this site, In the configs hide sensitive information such as passwords and public IP addresses and indicate which interfaces are for the host doing the copy and which interfaces are used for connection between switch and firewall and firewall and remote site.

HTH

Rick

[Larry Sullivan]

So if 1GB works fine, at what point does the 5GB file fail?  Does the 1GB max out the speed as well?  If so, for how long?

What is the wireless protocol that maintains connectivity between the two sides of these wireless devices?  Is the overhead for this protocol segregated from production traffic as it should be (might have to check with provider)?

To prove link stability in general, a simple ping test would due.  To prove stability when speed is maxed out, you would have to refer to the previous question about wireless protocol overhead.  

Your provider has probably dealt with these types of issues many times and should have a general explanation for you.