Network resiliency on flaky internet
Our customer has Velocloud SDWAN on raw internet from Middle east to US. The internet does not have dedicated BW. An internet link will have several TCP sessions on it, but once in while the available Internet bandwidth drops down significantly causing queuing delay and packet loss for these sessions. Since bandwidth is not guaranteed leaving Middle East, this causes TCP sessions to stall and sometimes drop off. While SDWAN can link steer onto a better links on TCP reset this does not work well for an SSH session copying a file. Is there a way to issue like this w/o getting a new MPLS link. Please advise.
There was question very similar that Aaron Tomosky had answered before, as below.
https://www.experts-exchange.com/questions/29142968/Ideas-on-how-to-cutover-to-a-backup-line-when-primary-goes-down.html
There was question very similar that Aaron Tomosky had answered before, as below.
https://www.experts-exchange.com/questions/29142968/Ideas-on-how-to-cutover-to-a-backup-line-when-primary-goes-down.html
David Favor
Using rsync may be the simple solution, as rsync is far more intelligent than scp + can resume transfers for connections that drop completely.
ASKER
Is rsync is a separate product? Haven't heard of it
hey Totaram. if you are sending your traffic via the overlay, velo should handle this for you. so are you going through a hub site, a velo gateway, or actually out the bare internet?
ASKER
Middle East sites are brach sites, sdwan controller is in US.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
Does it still work for file copying on SSH Session? I have told that the Normal users don’t even notice the issue (hiccup), but developers do.
ASKER
Hi ;
I saw an article on reddit talking about the issues w/ velo DMPO limitations
as below:
https://www.reddit.com/r/networking/comments/5gyem6/velocloud_claims_snake_oil/
I saw an article on reddit talking about the issues w/ velo DMPO limitations
as below:
https://www.reddit.com/r/networking/comments/5gyem6/velocloud_claims_snake_oil/
if there is only one circuit, and that circuit stops passing traffic, I suppose ssh file transfer could timeout. but regular old packet loss should not exist on an overlay tunnel. happy to review your config, sounds like something isn’t setup right. also it would help if you can identify exactly when this has happened so we look at the logs, or even better, a way replicate the issue,
also, suggest you watch this series on how the different parts of DMPO work, it will help this all make sense. some stuff works on one link, and some features obviously require another active path.
https://www.youtube.com/playlist?list=PLbssOJyyvHuVt4Pnn8SSFL2cCN-TKgTJk
https://www.youtube.com/playlist?list=PLbssOJyyvHuVt4Pnn8SSFL2cCN-TKgTJk
ASKER
Allow me some extra time, I will come back after investigation.
ASKER
I just learnt that customer is using some direct traffic which may not go through VCG. Is there a way to fix BW issues on direct internet connections, like P2P etc.
well, for velocloud to be able to remediate with FEC and jitter buffering and move packets around links, the traffic has to go over an overlay tunnel. basically be bookended with either a hub, another branch or the VCG for internet bound traffic. by sending traffic direct out the bare internet, you are limiting the ability of the platform to fix these issues. I'm going to need a diagram and some specifics of the config to help you fix this, but it should be within the capabilities of the platform to work much better than it is today.