BuyVM vs RamNode for GRE Tunneling

Discussion in 'Hosting Advice' started by dillybob, Jun 15, 2016.

  1. Hey guys,

    I found an epic deal for a dedicated server on LowEndTalk. From someone that is pretty reputable :)

    Only one problem... They don't have DDOS protection so I want to buy BuyVM or Ramnode and setup a gre tunnel :D

    Second problem.. How much ram do I actually need though? I know I need to buy one closest to the dedicated servers's location to minimize latency.. But, I am just unsure the performance issues around tunneling.

    Does GRE tunneling eat up alot of cpu? ram? or just bandwidth? :p Thanks!

    Edit: This is for around 1000 + players. (Not using Minecraft, but nodejs)

    Edit2: These players will not be really active though. Mostly idle, sending messages once every 20 seconds or so. Not a lot of usage if that helps
     
    #1 dillybob, Jun 15, 2016
    Last edited: Jun 15, 2016
  2. MikeA

    MikeA Retired Moderator
    Retired Benefactor

    With a GRE tunnel you'll need to consider the bandwidth it'll be using. You'll likely have high overage fees on any VPS that's being used for a GRE to a dedicated server with 1000+ connections. Both are good, RamNode has a great reputation and I'd highly recommend them. BuyVM is good but I haven't used them, but unless you're in a USA, UK or Canada they will likely blacklist you or ask for photo ID to let you order as they're very harsh on "potential fraud countries" (I've had numerous people come to me because they didn't let them order)
     
    • Like Like x 1
  3. Hey Mike, thanks!

    So basically, the 128mb vps will probably be pooped on? Maybe just find a dedi with ddos protection? :)
     
  4. MikeA

    MikeA Retired Moderator
    Retired Benefactor

    Well, not really, if it's just for tunneling traffic then it would probably work. I'd be more concerned about the usage for bad traffic that isn't mitigated immediately. There's tons of affordable dedi providers out there, so I never understand why people even bother with GRE tunnels anymore myself. I have clients that use them and oddly enough they mostly tunnel traffic to other VPS providers (??)

    Psychz, OVH, ReliableSite, are providers who are relatively cheap (<$100/m and you can contact them for a better deal most of the time) that offer a fair amount of mitigation. I use Psychz and OVH for my services right now and I never have issues with attacks, ReliableSite also has a good reputation but I don't use them. Adding a GRE tunnel between services just adds a point of failure and makes it a tad more annoying to troubleshoot network issues on either end of the point.

    just my 2c though, I never use tunnels.
     
    • Like Like x 2
  5. Yeah, I hear ya. I am just scared of those boot loading sites. My friend on discord has access to one and he doesn't even know how to install ram on a computer, and he took down one of my VPS's out of no where with a click of a button. It's just sad really. That's why I always want to be protected, but I am trying to find someone that can withstand those stupid bootloaders sites. I really like OVH and have been reading about them on these forums and I've heard they are not that bad and pretty good.

    Only one problem though, their server location is way too far east. I guess, I could use them for a UsEast server realm, but, I really am looking for a central server splat down in the middle. Texas / Kansas City area... very hard to find. I'll look into Psychz and ReliableSite.

    Also, the reason I am tunneling is because I found this retarded deal on LowEndTalk with a Xeon 2gb Server from "dediserve" (not sure if you had heard about them) -- but it's $7/month. It's a steal. Good location too, but then again, not in the same location as BuyVM's vps, so if I were to tunnel, the latency would be retarded. Anyways, thanks again for your help haha
     
  6. MikeA

    MikeA Retired Moderator
    Retired Benefactor

    Latency is really overrated. OVH's Montreal, CA datacenter is perfectly fine for all of the U.S., I think it's around 80ms in SoCal, I get 65~ms from central Texas and I never have issues with latency when connected to servers there. East coast is probably much less. Kansas City and Texas generally don't have many good DDoS mitigation providers who can handle game servers very well except Psychz probably. Anything you find on LowEndTalk, well, is probably going to be shit for hosting game servers.
     
    • Like Like x 1
  7. Well, reliable sites seems affordable at start, but at the end when you finish setting up the specs you want for your dedicated it will end having expensive costs as any other provider
     
  8. MikeA

    MikeA Retired Moderator
    Retired Benefactor

    RAM is typically one of the most expensive components, so of course.
     
  9. Yeah this is really tempting... I just live in WA State so that feeling of "server is being across the damn country" just makes it so unappealing to me! But then again, the players who do live in east coast are in love haha. I guess I gotta give up something
     
  10. Especially the 16GB modules and don't get me started on the 32GB modules...
     
  11. MikeA

    MikeA Retired Moderator
    Retired Benefactor

    Well, Psychz in Dallas would work if that's what you're worried about, but I've personally had no issues with OVH in Montreal. Latency from Montreal, CA to London last time I checked was only 95ms~, so it's a good location if you have a split amount of players in the U.S. and western Europe. But like I said, I don't personally see a little bit of latency being a big issue for the majority of players.
     
  12. Yeah, our client side prediction is pretty good within the game (nodejs, not mc one) -- so movement is done locally first and sent off. so the latency I guess isn't really going to be that big of a deal. And I never thought how close it was towards over in the WEurope area.

    I'm actually going to buy the ssd package 1 right now and test it out. There's also a 30% off coupon for the 6th and 12 month plans.. which is so tempting, but I want to try em out first just because. I think you're right and latency is overrated tbh now that I think of it.

    I'm just worried about those boot loading sites now. I'm pretty sure some script kittie will take it down that spends 3-5$ a month on those ddos sites but oh well... I feel confident OVH will mitigate most of it...
     
  13. $3-$5/month on "DDoS service" will do nothing to be honest. Basically the way DDoS works is a couple of machines are used to attack the target.

    For example at LEAST 10Gbps mitigation is not hard to come by usually and thus the attacker must have more throughout than that to flood it. 10Gbps will cost in the $x,xxx solely for the throughout. Stack on bandwidth charges and the price really get prohibitive.

    What you will most likely get hit by though are belief DDoS attacks which lasts a few hours at most. For smaller servers this won't be worth mitigating (because the down time won't be as disruptive) but larger servers should have at LEAST 50Gbps to do the job. As the attacks become frequent (up to multiple times a day for an "attractive" one).
     
    • Like Like x 1
  14. MikeA

    MikeA Retired Moderator
    Retired Benefactor

    If you want PM me, I probably have an empty server I can let you test on for a bit before I remove it.
     
    • Like Like x 1
  15. Yeah, that's why I'm so freaking worried and Cloudflare now supports Websockets.. But, they actually proxy all the packets, just not the initial connection :( So I have a feeling the latency would get worse.

    There is a game called wilds.io that just became popular, the dev is using a digital ocean approach where he just buys the droplets and the initial game server just acts like a manual load balancer and throws the new players on. I mean, obviously they havn't been ddos'd that much because they still got an active amount of players online and the guy would have to ddos around 10 or so droplets (I've talked to their dev and that's how many he has) -- but I mean, I'm not saying that's a solution but it could help mitigate. But then the problem is, since I use a central "server", if you will, if they target that one, I cannot do my manual load balancing.. and boom, it's over.

    I am eager to know if BuyVM or Ramnode can mitigate those "web based bootloaders", though, because that would be nice. I mean honestly, if someone wants to take down our game server, they probably will LOL. And I mean, there is really nothing to stop it so just prepare for the worse and mitigate it as much as possible. But I do think protection from those "bootloader" sites should be available from some hosts.. I hope :(

    Edit: Massive typo's sorry about that
     
    #15 dillybob, Jun 16, 2016
    Last edited: Jun 16, 2016
  16. No problems about hooking up additional Bungee instances solely to "mitigate" DDoSes may not be effective (unless spammed like that dev). But you should keep that in mind if you have quite a large player base. I believe people recommend something like an instance per 500 connections. To optimize performance and at least have a slight chance of keeping a partial network online.

    In my opinion go somewhere like Reliable Site where they can hook up with massive mitigation when the time comes. So when it needed you can ask them to "crank it up" receptivity to prevent adverse effects in the future.

    With the adove in mind I would start at say 10-20Gbps for a small server and at least 50Gbps for a large setup. Then should you starts to get attacked regularly above the purchased amount crank it up to the attack reported amounts and then some to keep your servers online reasonably (it won't be "bulletproof" but unless you got money to blow it only worth mitigating when you suffer some disruptive downtime).
     
    • Like Like x 1