</p> We take a very in-depth look at what it takes to power Quakecon. Qain gets nerdy in the NOC.
OpenBSD, PF, nginx, and several other interesting things are being used to power the network operations center at Quakecon. Check out the video for all the details.
the whole Lancache is described in detail here.
LANcache – Dynamically Caching Game Installs at LAN’s using Nginx
Have it running for LANs here as well saves so much bandwidth.
This was awesome!
Approximately how many people attend this event?
I think they said total attendance was over 7000 this year, but BYOC was somewhere between 2600 and 2800 I think.
EDIT: Correction, total attendance was recorded at about 11,000.
Cool, thanks!
Something to remember for the UK Lan @Kai @anon43112114
Cool stuff, that was a fun video
288 ports
72,000,000~ packets an hour
Just because.
PF tutorials on the tek would be great! Just a suggestion ;)
Hey did they ever figure out what was causing the down speed from the Steam cache to run at 100 K/s? Seemed like by Saturday everything was starting to run decently.
nice to see those things on the internet, I'm a network engineer and I always had a hard time explaining to people what I do in my job... the difference is that I have like 20 setups like this or bigger on my management lol...
The guy said that he used the same PF setup on OpenBSD (instead of pfSense) at home. In a home environment, unless you're really generating a lot of traffic, how would that queuing system really perform better than pfSense?
I didn't know that Squid could work just that quickly. It's also pretty genius to route the traffic to Nginx for caching and then have the cached data be served directly by Nginx later. I wonder if the HTTP headers get modified though (Server, Last Modified, etc.).
Also, I thought that Windows Update was encrypted (HTTPS)? How would they be able to cache that without enforcing a local certificate upon everyone and intercepting the encryption? I'll have to run Wireshark later and she if Windows Update really is encrypted or not.
AT&T 2 Gbps? Pathetic. They should have looked into this: TWC Dedicated Internet Access. Then you'll get 10 Gbps. I'm pertty sure that it's in the area. @wendell or @Logan please mention that TWC has a fiber line already on the Tek. They're just now rolling out mere 300 Mbps still over DOCSIS 3.0 in some areas (including Dallas).
Cool Video with cool people and cool tech!
Even though I am not very learned in the dark arts of network design and management/sysadmin, I still got a feel for what was in there and what got it to tick, well placed leading questions @DeusQain.
mo' please :)
Great video but I'm wondering if anyone knows what the requirements for the web-proxy (squid) were. From skimming through the Lancache description posted by @Hakker I gathered DNS was used to point the clients there instead of to the service provider (EA, Steam, etc).
I'm however wondering if clients had to manually set the proxy to whatever server was running squid or if there might be some automated way to do this in a BYOD/LAN type environment?
Personally I never see the real benefit of Squid on dynamic pages anymore. The amount that gets cached is nothing compared to what a LANcache will do. All pc's have local caches anyways so most images and such are already cached before they come to a LAN. Also I found it to have more drawbacks than just saving bandwidth as the cache doesn't update as frequently as most would want.
If people went on to go completely mad on Youtube we announced it if they kept going we just null route Youtube and the likes. One guy went screw you ppl so we set his speed back to a 14.4 modem. seriously you announce that the bandwidth is limited and if you get idiots who go youtubing all day long then stay at home if you are that addicted. A LAN event is like it says LAN (Local Area Network) party and not a WAN party. And yes even at LANS 2 places will screw you over and are a pain in the ass to cache. Youtube and Facebook.
@caprica The firewalls probably made it so that the DHCP configuration pointed the users to the proxy. Either the firewalls completely funneled the traffic that way (instead of to the internet) or the proxy was set up as the Gateway.
@Hakker I think that the cache was intended more for game downloading than other web stuff. Many websites are optimized to load quickly and have fewer and smaller resources that will have simple load spikes, but game downloads are just raw downloads of several gigs. If TF2 has a 1 GB update, then it can be downloaded once and dispersed from the cache server after that to save bandwidth.
But then you wouldn't need squid so they cache webtraffic too. I haven't done that for LANs in years. Basically stopped doing it since it became more dynamic content than static.
They use Cisco WCCP (Web Cache Communication Protocol) the core switch can detect any packet heading to the internet using tcp/80 and redirect the request to the proxy/cache mechanism, so the client never know that the connection is passing through a proxy. No setup needed on client side.
I too would enjoy a Tek Syndicate PF tutorial.
For the record, one of those 48 port Blades, was for backup only.
So it was basically empty.
Also, the core switch, they increased the PSU to the beefier supply. to handle the extra ports over last year.
Interesting video, thanks! What a nice setup; who is organising all of this? I'd like to see that guy's home setup next.