Solution for noisy network that corrupts downloaded files

Question

The network on the company where my wife works has issues. Almost every single file bigger then a few hundred kB is corrupted during download and even web pages are sometimes garbled. We checked all we could possibly check and it really isn't a problem on her computer (specially cause every other computer terminal in the building have the same problem).

All files are corrupted, including documents, photos, web pages... this happens for files downloaded through any means (various browsers, wget, download-managers, ...) except Dropbox. but the worst problem is that she can't install anything in her Ubuntu machine because the packages are corrupted at download and there's a checksum mismatch every time she tries to use apt-get or even run a downloaded script.

The network admin doesn't seem to be interested in correcting this and he doesn't even seem to understand what the problem is and what can be causing the problem (he insists that this is caused by heavy traffic in the network).

What is strange is that the files downloaded by dropbox are not corrupted. I think maybe it uses a checksum test and tries to download again in case it fails.

So, we thought that maybe there's some program that makes some kind of check during the download and download again the corrupted bits? Is there any way we can use this noisy network?

There's really no other option, it's the only fast network she has access to... :(

EDIT: I'm even losing ssh connections over this network. Can't manage to stay connected via ssh for more than 30 seconds... :( I get a:

Corrupted MAC on input.
Disconnecting: Packet corrupt

Answer 1

TL;DR: Sounds like Internet connection oversaturation. Try off-times, torrents, download managers, anything able to redownload only damaged portions of files. Also some network tips at the end.

I'll spare you most of the scolding for what sounds like your wife making personal use of the network at her office (/me slaps you on the wrist, consider yourself scolded), but if an office is the only available high-speed connection it makes me think you're in an area with generally poor broadband availability.

If that's the case, it's quite possible that the business connection in question is a T-1 line or perhaps bundled T-1s for higher capacity (or even a satellite link). Any of those options generally give performance that is at the low end of modern broadband speeds, and it's quite possible that the external network connection is saturated. Heavily oversaturated connections can easily lose packets, at least as far as the applications expecting them are concerned, because by the time a missed packet is re-requested and delivered the application has given up expecting it.

There's not going to be much that you can do beyond using block-oriented download methods such as BitTorrent (and Dropbox) that do checksumming of each block and which can transfer only mismatched blocks to reduce network traffic. Rsync (possibly tunneled over SSH) is another method to transfer files with that kind of block-level checking. Your best bet (at least for file-based transfers, not much help for email/browsing) may be to get another account outside the network where you can download files and "stage" them either into your Dropbox account or so you can then use rsync for downloads.

If you can get the network manager interested (assuming he hasn't already done some of these things) there may be ways to reduce the demand on the network connection such as implementing a transparent caching Squid proxy, banning/blocking streaming & torrents, and adding "greylisting" or an email spam filtering system that includes it between your mail server and the Internet (assuming that email is handled in-house).

Edit: Squid proxies and greylisting can actually be surprisingly simple to set up, particularly if you're using VMWare. Prebuilt VMs with Squid and filtering are available, and for greylisting ESVA (www.global-domination.org) is a decent choice if somewhat neglected these days.

Answer 2

TCP/IP has got checksum automatically. If there is corruption, it sends reset packet and all starts over again

we used to have oversaturated internet line all the time but underlying HW always worked flawlessly.

it is hard to pinpoint what is wrong unless I see how is all connected, set-up.

Change the job :-P

Answer 3

My guess is there is file inspection happening that is causing the corruption. Maybe there is a Anti-malware gateway on the network and perhaps this service cannot inspect SSL sessions (would explain Dropbox success).

I do not know whether apt-get does or can use SSL connections, but its worth a look.

Answer 4

You could use tcpdump or Wireshark or some other package to trace the network traffic.
(Note: 3rd party compiled package needed for Wireshark on Ubuntu).

This could help you figure out what is going on.