Best practice network model for 350~ lanparty model
-
Hello everyone
Our lanparty is growing fast and we want to get our internet access in a good shape for our users.
We have to work with what we have. So let me explain:We have 3x 150 Mbit internet connection.
We have a relative small amount of hardware to share them (old proliant servers).
Our network is Layer 2 it supports VLANs but no routing so that's not an option.
We work with an /16 network segment. More than enough IP's for every participant.One of the things we have in mind is to create 3 separate gateways, and divide the participants manually over those gateways by granting 1/3rd of the participants an other gateway IP.
Of course, there is no failover on the gateways then. We worked this way last year and it worked pretty well, but not perfect.We didn't make any usage of squid caching. We will try this out this year for the first time.
What is your oppinion on this situation? How would you resolve this? What hardware should you have in mind? Please keep in mind that it should be a low-cost solution.Thanks for your input!
Kind regards,
LP -
If this is for gaming, I wouldn't run squid necessarily - I don’t think it would do anything for you.
-
But wouldn't squid help lower the http traffic? (youtube / ustream / steam / ..) so there will be more bandwidth available for gaming?
-
Thats all dynamic content, so I doubt it. It will accelerate basic web browsing minus alot of dynamic audio/video content. Its actually damn hard to get really good use of squid for dynamic audio/video. Some systems claim to do an OK job but I don't know anyone who is using it well and successfully that way. If you were using squid to try to cut down on bandwidth because you were being charged per megabyte or something, it might save you 5%. I do use it. 5% cache hit is about what I get.
-
OK Thanks kejianshi, I'll take notice of that. Is it OK to split the 3 WAN's into 3 separate gateways with each 1 pfsense server to share it? Or is 1 server with 3 WANs a better solution? Someone told me you should never put more than 200 clients on 1 gateway. Of course, low latency is a must.
-
I would do Mulit-WAN load sharing with round-robin. Most people don't get the best use out of Multi-WAN because lots of them are trying to aggregate bandwidth to a single user, but you are not so you should get big bang for your buck.
-
I would do Mulit-WAN load sharing with round-robin. Most people don't get the best use out of Multi-WAN because lots of them are trying to aggregate bandwidth to a single user, but you are not so you should get big bang for your buck.
If you do go this route, don't forget to set sticky sessions, otherwise some games will get pretty sad if stuff is coming from multiple IPs.
-
Yep - Thats a check box.
-
But wouldn't squid help lower the http traffic? (youtube / ustream / steam / ..) so there will be more bandwidth available for gaming?
It will to a certain extent but Squid can cause problems for online games that authenticate through http but not for the actual game traffic (Many F2P games do so and I've had it break Blizzard Battlenet as well).
Furthermore, you will only save the bandwidth if the data set is repeatedly accessed.