high end server in a data center still chokes?

[windle]

Can anyone shed any light on this:

2x Quad Core processors @ 3.0GHz
4 Gb Duel Ram
2x 15,000 SAS Drives (raid 1)
100 meg unlimited connection in a datacenter. Speed test results in:

80 Megabits down
60 megabits up

Server averages about 3.5 meg upload so loads of scope for bandwidth?

CSS server is 26 man @ 100 tick (2 reserved so generally 24) is it just the who dont run over 28 man at 100 tick or is there some reason what causes this? Rcon stats reveals

0.00% CPU load FPS never drops below 512 yet in game as a client it looks like it does? but some maps more than others cause quite bad choke. ? Reduce tickrate or anything else to look at ?

[Derek]

Choke is almost 95% of the time not a server side problem. You are also running a 26 slot server @ 100 Tic which is really pushing it.

[m0-]

DerekRDenholm: Uhm I don't believe it is the slot issue as well, According to his box, he has 2 Quad core processors, therefore he has 8 cores, so he could put at least 200 slots on that server since (of spanned across all 8 cores). So obviously 1 server with 26 slots at 100 tick is enough...

When you mean choke, most likely it is a datacenter routing issue. Check your MTR settings to check which hop is loosing packets. I had the same issue with my data center and I while playing on my Server, the users and I ran a MTR test to that server. And found out some hops were packet lossed causing the choke. Send that MTR log(s) to your data center and they will open a ticket with the carrier to resolve issues. Might take 3 days to a week for them to find out what is going on.

For windows: WinMTR
For linux: MTR

While playing, run that MTR to that server ip, and hope choke occurs so that you can see which hop it occurs.

Good luck

[Drocona]

What DerekRDenholm was pointing at is the amount of players in a single server running 100 tickrate.

We don't recommend anything higher than 18 slots in a single server of 100 tick, going above this amount of player needs quite some experience to get right.

I personally think your problem is right there, try setting it 66 tick and see if the choke disappears (66 tick should be fine up to 32 players)

[windle]

i agree on the tickrate just checking realy as FPS and CPU are very stable and hardly being touched at all. The line is 60 meg up just sondering more than anything else i guess?

[Drocona]

If the FPS is stable (no matter how high, as long as it's higher than the set tickrate) and the Core usage is below 80% there shouldn't be a problem at all.
Something else you can do is do a traceroute and see if you have any loss or "hops" that aren't responding as they should, those can magically cause choke too (without loss going up....)

[windle]

tickrate has never dropped below 500, cpu load i give it access to the 8 cores and as a result it always reads 0% no 1 individual CPU core goes above 40 though

[Drocona]

I suggest always assigning it to 1 core, gives better results (since SRCDS uses only 1 core at the time, better give it 1 dedicated than being switched over the cores when it gets autobalanced.)

[m0-]

Once again, check your MTR. I had the same issues with choke and that is really the carriers. So check which hop causes the loss.

Your server is more than fine, and if your running a 100tick config, that doesn't make sense of the packet loss happening, cause choke is dependent on some loss factor.

http://www.google.ca/search?q=define:choke+packet

[windle]

i have tried running the MTR software but seams to just hang, i am on vista perhaps its for XP only? CPU seams to work very well giving access to all 8 all 3 servers seams to mainly use 1 core on their own anyway and always have access to another 7 if needed he he

[windle]

have booted a virtual machine up on xp and now works so i guess vista is a no no, anyway with all servers empty these are the results: which look fine apart from hop 10 presume its the final hop to my server and ping is not enabled. so i need to run this when the server are full and choking to see where the problem lies?

|------------------------------------------------------------------------------------------|

| WinMTR statistics |

| Host - % | Sent | Recv | Best | Avrg | Wrst | Last |

|------------------------------------------------|------|------|------|------|------|------|

| 172.31.75.254 - 0 | 9 | 9 | 0 | 2 | 10 | 0 |

| 10.87.88.1 - 0 | 9 | 9 | 10 | 28 | 160 | 10 |

|oldh-t2cam1-b-ge92.network.virginmedia.net - 0 | 9 | 9 | 10 | 11 | 20 | 10 |

|manc-t3core-1b-ae2-0.network.virginmedia.net - 0 | 9 | 9 | 10 | 12 | 20 | 10 |

| man-bb-b-as0-0.network.virginmedia.net - 0 | 9 | 9 | 10 | 12 | 20 | 10 |

| man-bb-a-ae0-0.network.virginmedia.net - 0 | 9 | 9 | 10 | 11 | 20 | 10 |

|nth-bb-b-so-300-0.network.virginmedia.net - 0 | 9 | 9 | 10 | 17 | 30 | 10 |

| tele-ic-1-as0-0.network.virginmedia.net - 0 | 9 | 9 | 10 | 16 | 21 | 10 |

| 212.250.14.86 - 0 | 9 | 9 | 20 | 22 | 30 | 20 |

| No response from host - 100 | 9 | 0 | 0 | 0 | 0 | 20 |

|________________________________________________|______|______|______|______|__​____|______|

WinMTR - 0.8. Copyleft @2000-2002 Vasile Laurentiu Stanimir ( stanimir@cr.nivis.com )

[Drocona]

Yes, windle, when you see choke run the traceroute.

I've had various people complaining about choke on their server, when I did a traceroute when they said choke started I'd see quite some loss at one or two hops.

About the CPU, the SRCDS can NOT use multiple cores. It uses 1 and 1 only.

[css]

windle Wrote:have booted a virtual machine up on xp ...

What virtualization software it's running? My guess is the choke is caused by the virtualization.

Do you really have the whole 2xQuad core server system for yourself or do you just have one VPS off the whole system, but they just sold it for you as 2xQuad system?

[windle]

no virtualization and yes i have the lot to myself

i ran the choke test when i was choking, saldy from another machine (virtual on same comp though) this is what i found:

|------------------------------------------------------------------------------------------|

| WinMTR statistics |

| Host - % | Sent | Recv | Best | Avrg | Wrst | Last |

|------------------------------------------------|------|------|------|------|------|------|

| 172.31.75.254 - 0 | 2850 | 2850 | 0 | 1 | 20 | 0 |

| 10.87.88.1 - 1 | 2850 | 2849 | 0 | 12 | 181 | 10 |

|oldh-t2cam1-a-ge92.network.virginmedia.net - 0 | 2850 | 2850 | 0 | 14 | 160 | 10 |

|manc-t3core-1a-ae2-0.network.virginmedia.net - 1 | 2850 | 2849 | 0 | 13 | 131 | 10 |

| man-bb-a-as0-0.network.virginmedia.net - 1 | 2850 | 2849 | 0 | 13 | 90 | 10 |

|nth-bb-b-so-300-0.network.virginmedia.net - 1 | 2850 | 2848 | 0 | 18 | 170 | 10 |

| tele-ic-1-as0-0.network.virginmedia.net - 1 | 2850 | 2847 | 0 | 19 | 140 | 20 |

| 212.250.14.86 - 1 | 2850 | 2848 | 0 | 95 | 4877 | 20 |

| No response from host - 100 | 2850 | 0 | 0 | 0 | 0 | 0 |

|________________________________________________|______|______|______|______|__​____|______|

WinMTR - 0.8. Copyleft @2000-2002 Vasile Laurentiu Stanimir ( stanimir@cr.nivis.com )

[windle]

i presume that :


| 212.250.14.86 - 1 | 2850 | 2848 | 0 | 95 | 4877 | 20 |

4877 at worse point is not very good!

but in general most hops lose it seams! normal or high?