Jump to content
[DE]Drew

Dedicated Conclave Servers

Recommended Posts

Totally agree with iseeu2 as commercial servers are less conventional due to security issues.

A "hack around" with teamviewer or similiar solution is not suitable.

AWS g2.2xlarge instance is not cheap too.

Share this post


Link to post
Share on other sites

Just saying that the way you go into the servers is mostly for the virtual server itself not the rest, you can keep the box you run it in just in control with team-viewer.

I just read that he was going to use a AWS just used that as a example, I get the security issues but just for one box it is not as a big problem.

My screen was not totally explaining but that was the top end of my server, you just could go into the windows box and install some team-viewer client in there and monitor it from there, without interfering with the top system.

Saying this the top system that has to be secure, if you just give it internet access only and no local network access it is a lot more secure and probably not even as hack-able as you guys say.

I use it to remote control my server and or boxes and it works fine so far and it is just safe enough for most jobs, team-viewer has a history of being hacked but that is fixed so far.

Anyways it is your own choice, also multiple server hosting so far is impossible because the servers don't support this (Aside from the fact that you can have multiple accounts, but just do not do this).

 

Edited by Nep_Blanc

Share this post


Link to post
Share on other sites

My point is not as much from a security viewpoint as it is from a convenience one. Warframe's launcher detects terminal services as an attempt to launch it remotely, and disallows that.

Teamviewer, as I mentioned earlier, is good for conventional use. But there's a reason it's not used for production environments on Windows Servers. One, you'll have to hack your way through system edits to make it work to begin with, and two, it's not a core service so it's not reliable.

How it works on an OS built for desktop use vs how it behaves on an OS built for server use is very different. Unfortunately, AWS instances (and servers in general) do not deploy Windows 7/8/10, rendering Teamviewer use more or less a moot point.

If DE really wishes to 'expand' on player based server hosting, they need to consider:
- Allowing the console to be launched remotely, independently of the launcher, and not require gfx cards.
- Letting users fire up multiple instances, even if this doesn't credit the users with 'karma'. Frankly, I don't care about some points on a webpage. I'm guessing most users won't, either.

And in the long term, they need to figure out a way to let the users host servers on Linux variants without relying on WINE. WINE is not feasible in the long term.

Share this post


Link to post
Share on other sites

Hello DE!

1) Can you please optimize dedicated server so it takes a few RAM and disk space(server does not need any GUI or rendering stuff so it can be cut out)?

2) Can you please build native static Linux version of server?

3) Is it possible to have a few plat for running a server 24/7(like 1-5 a day if server is widely used)?

If all 3 would be done i would enjoy hosting several servers on my hardware that is mainly used for other things like compiling software and network research.

i think 5400mbit(my summary bandwidth across Europe) would be more than enough to run several servers for people, even under intensive load ;)

Share this post


Link to post
Share on other sites
On 2016-11-09 at 3:32 PM, Death_Master_ said:

Hello DE!

1) Can you please optimize dedicated server so it takes a few RAM and disk space(server does not need any GUI or rendering stuff so it can be cut out)?

2) Can you please build native static Linux version of server?

3) Is it possible to have a few plat for running a server 24/7(like 1-5 a day if server is widely used)?

If all 3 would be done i would enjoy hosting several servers on my hardware that is mainly used for other things like compiling software and network research.

i think 5400mbit(my summary bandwidth across Europe) would be more than enough to run several servers for people, even under intensive load ūüėČ

Optimizing use for server usage is a thing I want too it is a thing and burden to sort-of accelerate for the launcher and that is totally unnecessary. Although having a UI for the server stats and other things like MOTD is a thing I Want to see in the next iterations.

Linux version is a wanted thing for me too because running a windows version in a virtual-box is not a most efficient ever but it works so far.

Platinum rewards are not a thing that we want I discussed this with other server owners and getting plat based on points is not the way to go, sorry to say but this is not the way to do this. they said they will reward us the server hosts for hosting but they will make a non saltizing way for it I hope because hosting servers and getting plat for it is kinda weird and pointless for DE.

I second with sufficient hardware you can host multiple matches with ease, so saying that PVE/PVP hosting with multiple running matches is a thing that can be realized in my opinion and help the community really well.

And well that bandwidth is enough to host your own private youtube service XD.

Edited by Nep_Blanc

Share this post


Link to post
Share on other sites

After discovering the dedicated machine I had set up had restarted, I haven't gotten a single match played in over 12 hours, nor have I seen any connection attempts. Anyone else been affected today?

Edited by StalkingyouGames

Share this post


Link to post
Share on other sites
9 hours ago, TheRoaringLion said:

Although having a UI for the server stats and other things like MOTD is a thing I Want to see in the next iterations.

That could be just simple webui(that won't take much resources to run and can be even optional :))

I'm pretty sure it can get as low as 20-30MB of RAM and a few cycles of CPU for a server if optimized fully. 

10 hours ago, TheRoaringLion said:

Platinum rewards are not a thing that we want I discussed this with other server owners and getting plat based on points is not the way to go, sorry to say but this is not the way to do this. they said they will reward us the server hosts for hosting but they will make a non saltizing way for it I hope because hosting servers and getting plat for it is kinda weird and pointless for DE.

I didn't fully understand here, sorry! To me a few plat would be good as it will pay me formas/reactors/catalysts i use. And if run long can even be enough to buy a frame or two from game shop(So instead of spending real money in game shop directly i will run part of game infrastructure for them(and community) which uses part of resources i pay money for). 

Share this post


Link to post
Share on other sites
On 2016-11-10 at 7:04 AM, Death_Master_ said:

That could be just simple webui(that won't take much resources to run and can be even optional :))

I'm pretty sure it can get as low as 20-30MB of RAM and a few cycles of CPU for a server if optimized fully. 

I didn't fully understand here, sorry! To me a few plat would be good as it will pay me formas/reactors/catalysts i use. And if run long can even be enough to buy a frame or two from game shop(So instead of spending real money in game shop directly i will run part of game infrastructure for them(and community) which uses part of resources i pay money for). 

Web UI could be awesome but ah well we will see how this ends up to be.

The platinum rewards are a no go, just to say you can like if you see the points now farm enough plat to not even have to pay the game anything, they have said to reward us, but I hope it is in a way that it is like syndicates.

Yes getting additional forma and or potatoes is nice but Let that be to the DEVS and the designers to decide to make something not only exclusive to the server hosting.
We pay for the power and the machines running but you gotta remember it is voluntary service so DE has nothing to owe you in any sense.

Edited by Nep_Blanc

Share this post


Link to post
Share on other sites
23 minutes ago, TheRoaringLion said:

We pay for the power and the machines running but you gotta remember it is voluntary service so DE has nothing to owe you in any sense.

 

Yep! I'm not demanding plat, i'm just telling i would like it as reward and only they decide what rewards they are giving(and even decide give rewards at all or not)!

And depending on what they wish to give i can decide how much(effort/money/resources) i wish to give ^^

Anyway on Windows platform it would be 1 or two servers, not more because i don't want to run VM for it and don't have lot of Windows machines.

Share this post


Link to post
Share on other sites
On 2016-11-10 at 5:34 PM, Death_Master_ said:

Yep! I'm not demanding plat, i'm just telling i would like it as reward and only they decide what rewards they are giving(and even decide give rewards at all or not)!

And depending on what they wish to give i can decide how much(effort/money/resources) i wish to give ^^

Anyway on Windows platform it would be 1 or two servers, not more because i don't want to run VM for it and don't have lot of Windows machines.

Although got to agree with the fact that Linux optimized servers with a possible UI on the GUI or a Web-interface with multiple sessions hosted is a better future but for now DE has to realize this and we don't know for now if DE is interested in this kind of services.

Its hard to say but most of the peeps that host a DD server can have a potato rig and not really be a good way to host, some have Premium rigs but have bad internet or just have both right but are neglected because of other reasons.

DE should better make a sort of Benchmark system to make the servers based on the performance and internet speed they have and based on that assign the hosting and or how many sessions you host and or if you can ever host a PVE match.

My server on the moment can possibly host 16 PVE matches and my internet can cope with all of it easily.
Based on that logic I can basically host 32 PVP matches and then I run into a bandwidth problem possibly.

For now I host my DD Warframe server in a Virtual-box and it works fine people enjoy it and seem to have fun on it.
 

 

Edited by Nep_Blanc

Share this post


Link to post
Share on other sites

Hi guys! I really appreciate all the experiments/ideas (AWS!). FWIW, server itself doesn't use GPU at all, it's 100% headless text app. Launcher is an obstacle here, but we'll try to work around that. Also, not sure if has been mentioned in the patch notes/FAQ, but as of TWW you can actually host servers on multiple machines and they can all be associated with single Warframe account at the same time (just need to run WF once on each machine, so it 'binds' to your account).

Share this post


Link to post
Share on other sites
5 hours ago, maciejs said:

Hi guys! I really appreciate all the experiments/ideas (AWS!). FWIW, server itself doesn't use GPU at all, it's 100% headless text app. Launcher is an obstacle here, but we'll try to work around that. Also, not sure if has been mentioned in the patch notes/FAQ, but as of TWW you can actually host servers on multiple machines and they can all be associated with single Warframe account at the same time (just need to run WF once on each machine, so it 'binds' to your account).

Thanks for the information.

Ping problem on DS: it is much higher than it should be. This is not normal, when you have 35-40 ping, on your own server, with 100Mbps internet (speedtest - 2 ping). Some players from other cities in my country, with the same internet speed, have more than 100+ ping on server. 

Edited by knife007

Share this post


Link to post
Share on other sites
On 2016-11-18 at 1:51 AM, knife007 said:

Thanks for the information.

Ping problem on DS: it is much higher than it should be. This is not normal, when on your own server ping 35-40, with 100Mbps internet (speedtest - 2 ping). Some players from other cities in my country, with the same internet speed, have more than 100+ ping on server. 

I've seen 30 ping on my own server probably the Proxy Warframe uses IDK but this does not seem to be needed in my opinion tho.
Less ping means better gameplay so if you remove that step that would make PVP a lot better too.

Edited by Nep_Blanc

Share this post


Link to post
Share on other sites
On 2016-11-17 at 8:02 PM, maciejs said:

Hi guys! I really appreciate all the experiments/ideas (AWS!). FWIW, server itself doesn't use GPU at all, it's 100% headless text app. Launcher is an obstacle here, but we'll try to work around that. Also, not sure if has been mentioned in the patch notes/FAQ, but as of TWW you can actually host servers on multiple machines and they can all be associated with single Warframe account at the same time (just need to run WF once on each machine, so it 'binds' to your account).

Ah well more Virtual boxes for me on my server I can run them and my server can take it so I am good with it.
 

Edited by Nep_Blanc

Share this post


Link to post
Share on other sites
On 2016-11-17 at 7:51 PM, knife007 said:

Thanks for the information.

Ping problem on DS: it is much higher than it should be. This is not normal, when you have 35-40 ping, on your own server, with 100Mbps internet (speedtest - 2 ping). Some players from other cities in my country, with the same internet speed, have more than 100+ ping on server. 

Part of that might be due to how we calculate ping internally. I'll take a look and make sure it's closer to reality (I want to stress - this doesn't affect gameplay, this is purely for informative purposes).

Share this post


Link to post
Share on other sites

OK, so back to the topic of ping. Seeing as we have a little bit of a technical crowd here I'll delve into some details to better explain why it's higher than expected. TLDR summary is: displayed value is not just the time it takes for the packet to travel from your machine to the host and back, it includes time it takes to process that data on both ends.

In order to optimize screen to screen latency, Warframe handles any new packets at the beginning of the frame, before frame logic, so that we can act upon on them immediately and sends responses at the very end, immediately after generating new data. We've done a few passes on this setup, trying to move these calls around to provide the best experience (still should have my high-speed camera captures somewhere when I record both screens and then count frames). We do not continually poll for new data, though, it only happens once every frame. Assuming a purely theoretical situation of 0 network delay and both client and host running at 60fps:

0) client sends a packet at time 0

1) it reaches the host immediately and we're lucky enough it's just before host starts his frame, he receives it, acts upon it and generates a response

2) host sends a response 16ms later

3) client receives a response, updates his ping measurements

As you can see, client will be led to believe his ping time is 16ms (even though the network transfer time is '0'). It's entirely possible our packet reaches the host when he's in the middle of his frame, so he'll handle it a little bit later, so 16ms is actually a lower bound. Now, it's not super difficult to work around some of these issues, e.g. we could inform the client that it takes us 16ms to process our frame, so he can adjust the results. You could also send the acknowledgment message immediately, without waiting for the end of the frame (this is what we do for QOS pings that happen during matchmaking, we do want this to be as close to 'network transfer time' as possible). In this case I do feel it's better to include processing time, though. This seems like a more truthful estimate of a 'game ping'. This is how long it takes for you to receive the response to your actions (host needs to generate it). To use an extreme example again, if you play a host with excellent network connection (0 'network' ping) and a terrible framerate (10fps), if I 'correct' it, you might be led to believe your "game ping" is 0 and your connection is excellent, while in reality it feels more like playing with a ping of ~100ms (that's how long it takes for you to notice the results of your actions).

Having said all that, while investigating I noticed a bug that'd sometimes result in a (reported) ping higher than in reality (~half a frame difference), this should be fixed in the next build. Another thing worth mentioning the DS 'frame rate' is capped at 60fps by default. You can change it by editing the LotusDedicatedServerConfig section in the DS.cfg file. Corresponding entry is App.DedicatedServerFrameRate (e.g. App.DedicatedServerFrameRate=100 for 100 FPS). I do not recommend going super high here as DS doesn't need to render anything, so it can get really heavy on your CPU (ie. if you set it to 1000 it might actually hit it).

  • Upvote 3

Share this post


Link to post
Share on other sites
3 hours ago, maciejs said:

OK, so back to the topic of ping. Seeing as we have a little bit of a technical crowd here I'll delve into some details to better explain why it's higher than expected. TLDR summary is: displayed value is not just the time it takes for the packet to travel from your machine to the host and back, it includes time it takes to process that data on both ends.

In order to optimize screen to screen latency, Warframe handles any new packets at the beginning of the frame, before frame logic, so that we can act upon on them immediately and sends responses at the very end, immediately after generating new data. We've done a few passes on this setup, trying to move these calls around to provide the best experience (still should have my high-speed camera captures somewhere when I record both screens and then count frames). We do not continually poll for new data, though, it only happens once every frame. Assuming a purely theoretical situation of 0 network delay and both client and host running at 60fps:

0) client sends a packet at time 0

1) it reaches the host immediately and we're lucky enough it's just before host starts his frame, he receives it, acts upon it and generates a response

2) host sends a response 16ms later

3) client receives a response, updates his ping measurements

As you can see, client will be led to believe his ping time is 16ms (even though the network transfer time is '0'). It's entirely possible our packet reaches the host when he's in the middle of his frame, so he'll handle it a little bit later, so 16ms is actually a lower bound. Now, it's not super difficult to work around some of these issues, e.g. we could inform the client that it takes us 16ms to process our frame, so he can adjust the results. You could also send the acknowledgment message immediately, without waiting for the end of the frame (this is what we do for QOS pings that happen during matchmaking, we do want this to be as close to 'network transfer time' as possible). In this case I do feel it's better to include processing time, though. This seems like a more truthful estimate of a 'game ping'. This is how long it takes for you to receive the response to your actions (host needs to generate it). To use an extreme example again, if you play a host with excellent network connection (0 'network' ping) and a terrible framerate (10fps), if I 'correct' it, you might be led to believe your "game ping" is 0 and your connection is excellent, while in reality it feels more like playing with a ping of ~100ms (that's how long it takes for you to notice the results of your actions).

Having said all that, while investigating I noticed a bug that'd sometimes result in a (reported) ping higher than in reality (~half a frame difference), this should be fixed in the next build. Another thing worth mentioning the DS 'frame rate' is capped at 60fps by default. You can change it by editing the LotusDedicatedServerConfig section in the DS.cfg file. Corresponding entry is App.DedicatedServerFrameRate (e.g. App.DedicatedServerFrameRate=100 for 100 FPS). I do not recommend going super high here as DS doesn't need to render anything, so it can get really heavy on your CPU (ie. if you set it to 1000 it might actually hit it).

Thank you for your help and information. Now ping is much better.

Share this post


Link to post
Share on other sites

So for the past few days I've observed that my game client and the Dedicated server instance have been freezing together for some time, with an error message complaining about unable to access the cache file because the game instance is conflicting with the dedicated server instance. 

I can send support the required EE & Dedicated server logs if necessary.

Share this post


Link to post
Share on other sites
On 2016-11-24 at 11:33 PM, maciejs said:

Having said all that, while investigating I noticed a bug that'd sometimes result in a (reported) ping higher than in reality (~half a frame difference), this should be fixed in the next build. Another thing worth mentioning the DS 'frame rate' is capped at 60fps by default. You can change it by editing the LotusDedicatedServerConfig section in the DS.cfg file. Corresponding entry is App.DedicatedServerFrameRate (e.g. App.DedicatedServerFrameRate=100 for 100 FPS). I do not recommend going super high here as DS doesn't need to render anything, so it can get really heavy on your CPU (ie. if you set it to 1000 it might actually hit it).

Just a question can you give A good example of the DS.cfg file with that  App.DedicatedServerFrameRate=[your framerate] because I understand partially what you mean but just to be sure I don't screw up some settings and make the game experience a lot worse.

Thanks already for the help and explaining how this works.

 

Edited by Nep_Blanc

Share this post


Link to post
Share on other sites

Hopefully hosts are going to be proud owners of hexa-core or above i7-s with Swedish internet.

Share this post


Link to post
Share on other sites
2 hours ago, TheRoaringLion said:

Just a question can you give A good example of the DS.cfg file with that  App.DedicatedServerFrameRate=[your framerate] because I understand partially what you mean but just to be sure I don't screw up some settings and make the game experience a lot worse.

Thanks already for the help and explaining how this works.

 

Chao, The Roaring Lion

 

Sure. If you open your DS.cfg file (this is in AppData Warframe directory), you should find a section named LotusDedicatedServerConfig there, ie.

[LotusDedicatedServerConfig,/Lotus/Types/GameRules/LotusDedicatedServerConfig]

Just add this below:

App.DedicatedServerFrameRate=100

for limiting to 100fps.

 

  • Upvote 2

Share this post


Link to post
Share on other sites
16 hours ago, maciejs said:

OK, so back to the topic of ping. Seeing as we have a little bit of a technical crowd here I'll delve into some details to better explain why it's higher than expected. TLDR summary is: displayed value is not just the time it takes for the packet to travel from your machine to the host and back, it includes time it takes to process that data on both ends.

In order to optimize screen to screen latency, Warframe handles any new packets at the beginning of the frame, before frame logic, so that we can act upon on them immediately and sends responses at the very end, immediately after generating new data. We've done a few passes on this setup, trying to move these calls around to provide the best experience (still should have my high-speed camera captures somewhere when I record both screens and then count frames). We do not continually poll for new data, though, it only happens once every frame. Assuming a purely theoretical situation of 0 network delay and both client and host running at 60fps:

0) client sends a packet at time 0

1) it reaches the host immediately and we're lucky enough it's just before host starts his frame, he receives it, acts upon it and generates a response

2) host sends a response 16ms later

3) client receives a response, updates his ping measurements

As you can see, client will be led to believe his ping time is 16ms (even though the network transfer time is '0'). It's entirely possible our packet reaches the host when he's in the middle of his frame, so he'll handle it a little bit later, so 16ms is actually a lower bound. Now, it's not super difficult to work around some of these issues, e.g. we could inform the client that it takes us 16ms to process our frame, so he can adjust the results. You could also send the acknowledgment message immediately, without waiting for the end of the frame (this is what we do for QOS pings that happen during matchmaking, we do want this to be as close to 'network transfer time' as possible). In this case I do feel it's better to include processing time, though. This seems like a more truthful estimate of a 'game ping'. This is how long it takes for you to receive the response to your actions (host needs to generate it). To use an extreme example again, if you play a host with excellent network connection (0 'network' ping) and a terrible framerate (10fps), if I 'correct' it, you might be led to believe your "game ping" is 0 and your connection is excellent, while in reality it feels more like playing with a ping of ~100ms (that's how long it takes for you to notice the results of your actions).

Having said all that, while investigating I noticed a bug that'd sometimes result in a (reported) ping higher than in reality (~half a frame difference), this should be fixed in the next build. Another thing worth mentioning the DS 'frame rate' is capped at 60fps by default. You can change it by editing the LotusDedicatedServerConfig section in the DS.cfg file. Corresponding entry is App.DedicatedServerFrameRate (e.g. App.DedicatedServerFrameRate=100 for 100 FPS). I do not recommend going super high here as DS doesn't need to render anything, so it can get really heavy on your CPU (ie. if you set it to 1000 it might actually hit it).

Just to clarify, is this tweak effectively tickrate?

Edited by Phasedragon

Share this post


Link to post
Share on other sites
6 minutes ago, Phasedragon said:

Just to clarify, is this tweak effectively tickrate?

Pretty much. It still only affects server updates and individual object properties might not be transmitted every frame (so if you run at 100fps, doesn't mean we send positions 100 times per second, we do update logic 100 times per second, though).

Share this post


Link to post
Share on other sites

@maciejs

500hz mode :D / 8 player lobby = 60% cpu load on haswell i5 desktop.

bie8JuQ.png

144hz mode / 8 player lobby = 45% load

9rhhb5r.png

Normal 60hz mode = 30ms ping

----------

I noticed a definite jump in responsiveness going from 60->144hz mode.

Edited by Pythadragon

Share this post


Link to post
Share on other sites
5 hours ago, Neah said:

So for the past few days I've observed that my game client and the Dedicated server instance have been freezing together for some time, with an error message complaining about unable to access the cache file because the game instance is conflicting with the dedicated server instance. 

I can send support the required EE & Dedicated server logs if necessary.

Yes, this would be very helpful. Could you please send us a support ticket with a copy of your EE.log + Server log?

Thanks.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×
×
  • Create New...