100 man game.

ah well that would certainly be an issue

and i've never had any bandwidth issues on my instance in the nyc2 location

Most likely you just never tested it with bandwidth-greedy applications. I'm checked instances in different data centers and connection speed always was pretty unstable. It's not so important for personal VPS or just easy site hosting, but for this game it is.

Still when server released I'll try to host PA games on both digital ocean instances and my Hetzner servers.

PS: I made tests about 4-5 months ago, not sure if situation improved since then, but before my "desktop hardware" server from Hetzner was way cheaper and have much better connection.

x'D

For the record, those are insanely expensive prices, especially for linux instances. Most of our server hardware is rackmounted. We only switch to elastic computing for overflow. Elastic is always much more expensive, though I'm not sure why DigitalOcean's appears to be so much more expensive than Amazon.

So the issue is one of balancing. We don't have a system for segregating the servers based on capacity. We'll have to implement that.

And yes this means that when you can run your own servers you can make games as big as you want.

Might be real interesting sometime to hear what it costs to run this game. I'm thinking of a distant future where players would just lease their own Amazon cloud space to host servers....I looked into that briefly one time for another game and couldn't figure out what it might cost, like even a rough ballpark.

I thought this was the issue. Though it still seems strange to me on how the server prioritizes games, the bigger server gets the majority of the processing, and the other games get pushed aside, yet nothing in place to cap the total capacities or remove them? Glad to see that will be changed.

But will there be anything in place if we start going above our own caps in custom servers? Will there be an ability to 'reserve' a slot when creating a large game and waiting for the other games to end before it actually initializes it, stopping other games from starting until the large match is created? this is purely for custom servers.

But glad to know that I may yet prove this game can run with 100 people Wrong Cat, we need to get on this.

Note: I'm really glad that the insane event at least got some great discussion going between the Developers and the Community. THANKS UBER!

It's not really strange. The load balancing just assumes 8 games of 10 people. If you run some crazy game it doesn't have a feedback mechanism to deal with that so bad things happen. Mainly issues with memory.

When you run your own servers you will be able to set them up however you want.

I'm pretty stoked that I figured that out from first principals.

Even if I forget about memory and thought the issue was bandwidth.

Not sure why you think they are expensive, maybe if your looking at the monthly prices but switching it over to hourly you can see the prices how amazon prices their stuff.

From here: http://aws.amazon.com/ec2/pricing/
And here: https://www.digitalocean.com/pricing

I have been looking at the on demand instances on amazon. Since that is what is equivalent to digital ocean.

And everything over on amazon is atleast double the hourly prices.

if you want a small server.

m1.small 1 core 1.7 gb of ram $0.060 per hour, digital oceans second plan is 0.015 per hour 1 core 1 gb of ram. The storage amount isn't really equivalent. But i still don't see unless your looking at spot instances, which is only slightly cheaper in some cases, and that isn't guaranteed. How you think its expensive.

If DO is expensive then amazon is insanely expensive.

200 player servers here we come

Thanks for the clarifying post! Your first post, based on my knowledge of our SoftLayer costs was VERY expensive, but looking through the link, yah, their pricing at least where AWS is concerned is quite a bit more competitive.

Certainly worth investigating!

How much fuss and bother would it be for (from time to time) a large group of players organise themselves together and Uber allocates only one instance to a particular server?

No matter how cool this would be, it's probably a better idea to wait until Uber allows servers to react to large games.

I wonder how long that will take. It looks like simple math, but as much as I know about networking it could take a day's worth of man hours.

Anyway first you have to start host games on Linux.

Is the current PA server windows only??

I sure there was post from somebody from Uber that they're build server for all platforms.

But I mostly sure that all UberNet servers running on Windows. You can actually check your game logs and see their IPs.

Server runs on all platforms, but it's not as simple as just flipping a switch to make UberNet run Linux based servers, as we have a ton of scripting and other critical services all tied together. However, simply making the server linux and leaving the rest the way it is would, presumably, be pretty easy.. if we could do safe, fast and secure connections between disparate providers, which seems.. less likely.

I do understand that you can't just go and change everything on backend and as William noticed few days ago "it works-don't touch it". I sure you have more important work to do than just spending time on backend.

To be fair I don't really get what you mean here.
If provider itself is trusted there nothing more you have to do because even openvpn can be configured in hour, but to be fair I don't see what you might need except openssh.

linUX ! linUX! linUX!