Quad Core Xeon 2.4ghz

[ColoDedi]

Afternoon All,

Im currently deciding what server hardware to buy for a new server that my company is rolling out. The server will be running SRCDS for hosting CS:S servers and HL2DM also some DoD:S

The spec of the server i have come up with so far is the following
1x Quad Core Xeon 2.4ghz
8GB DDR2 800mhz Unbuffered Non-ECC Ram
1x 750gb SATAII Seagate Baracuda
100mbit dedicated internet link.

How many slots should i be expecting to run on this server?
I thought i would get the views of people running SRCDS already to get a rough idea.

I thank you in advance.
Kind Regards Greg Langford

[Hazz]

Looks good, apart from I don't think 8GB ram is needed, 4GB should be fine.

[dualcore1289]

about 40 100tickrate slots per core, so roughly 160. . .

[janne]

dualcore1289 Wrote:about 40 100tickrate slots per core, so roughly 160. . .

I'd say max 24 slots per core and 2-4 gig ram is enough

[Hazz]

I would say around 40.....

24 seems a bit low for such a powerful processor.

[ColoDedi]

Is there a way to set SRCDS to use a certain core or does the O/S do that part of the processing?

Im planing on putting 8gb in as it will work out cheaper than me deciding at a later date to put 8gb in as there is a charge for the remote hands service at the data center.

Would 3TB of bandwidth a month be a sufficient ammount for say 140 slots in total?

[dualcore1289]

percisely. . . 40 max per core id say. . . you can set each servers affinity in your task manager. . . and 8gb is an overkill, i doubt you'll need 4gb for that matter. . . and if all 140 slots were full 24/7 3tb wont be enough, but i doubt itll be full 24/7. . . id say stick with the 3tb, and upgrade later once you see your getting close to it.

[m0-]

dualcore1289: how did you figure that out

Guys read this, http://whisper.ausgamers.com/wiki/index.php/Tickrate#Server_Bandwidth_Calculation_for_Dummies kinda nice explanation.

Wow a good server wastes a lot of money per bandwidth.

[dualcore1289]

I figured that out from drocona's previous posts. (thanks) lol. and yea the more slots a server can hold the more bandwidth youll need to host them slots.

[ColoDedi]

I have calculated 128 slots all full at the same time to be 3.75 mega bytes (yes i know the difference between mega bytes and mega bits) i mean mega bytes as i know the server is capable of that easily as i have done a speed test on the network at the datacenter and this gave me 5 mega bytes a second when downloading a test file with wget.

Im not sure a 1gbps connection would be required for say 150 slots? it would seem from my calculations that 100mbit connection would be plenty enough.

[blik]

I would say 160 at 66 tick - definitely not at 100 tick. This is roughly what we got out of a Q6600 2.4Ghz - without knowing the exact model number of the Xeon im guessing it's the same processor as the Q6600.

3TB will not be enough for 160 slots at 66 tick if they are full 24/7. But I highly doubt you will get anywhere near to 24/7 usage with an IP that isn't popular/known (yet?).

[janne]

These are readings from a cs:s deathmatch server having 24 active players playing on de_dust2.
server spec: Intel quad Q6600 2.4ghz, debian lenny 2.6.26.3-rt5, 4GB 800mhz ddr2 ram. (From what I understod the Q6600 is the same cpu as the xenon and should be comparable.)

Here you can see CPU usage, Net usage, no players etc.
Just do the math to figure the rest out.

(all players rates forced to 35000,100,100. No source tv)

Code:

CPU In Out Uptime Users FPS Players
70.00 87901.77 449117.69 287 6 982.32 24
70.00 88795.45 440320.28 287 6 981.35 24
67.00 90264.26 429176.91 287 6 982.32 24
67.00 92140.31 437090.38 287 6 989.12 24
66.00 97474.78 417352.25 287 6 990.10 24
66.00 94501.98 424850.03 287 6 888.89 24
67.00 92255.65 460944.31 287 6 986.19 24
67.50 94727.71 415532.56 287 6 988.14 24
68.00 97427.03 407705.44 287 6 979.43 24
68.00 97589.80 449011.50 287 6 977.52 24
67.80 96861.81 485318.78 287 6 992.06 24
70.00 97244.15 450179.09 287 6 906.62 24
70.50 96208.81 476752.38 287 6 982.32 24
72.00 99704.41 449496.00 287 6 978.47 24
73.20 101865.42 473774.34 287 6 979.43 24
75.00 99826.45 501649.44 287 6 980.39 24
75.00 100222.92 486620.38 287 6 971.82 24
74.67 98073.15 518652.81 287 6 981.35 24
74.00 95493.09 510716.72 287 6 981.35 24
74.00 94384.85 496034.59 287 6 971.82 24
72.00 95140.52 522564.19 287 6 988.14 24
72.00 97404.47 513634.09 287 6 981.35 24
72.00 99084.70 485689.97 287 6 992.06 24
72.00 101431.77 454745.81 287 6 982.32 24
71.75 101808.84 444390.34 287 6 978.47 24
71.75 101408.34 445299.31 287 6 97.47 24
71.75 100416.16 448641.69 287 6 991.08 24
71.75 99662.09 450473.62 287 6 982.32 24
72.00 99496.46 448217.81 287 6 968.99 24
72.00 100401.26 459748.88 287 6 987.17 24
77.00 99827.54 475141.94 287 6 979.43 24
77.00 97831.33 481879.50 287 6 986.19 24
77.00 97439.02 477639.06 287 6 992.06 24
77.00 96861.07 472707.75 287 6 971.82 24
74.50 96363.12 474293.06 287 6 981.35 24
74.50 95689.21 478704.19 287 6 981.35 24
74.50 95345.36 482105.16 287 6 982.32 24
74.50 94763.42 479498.69 287 6 990.10 24
74.50 93480.60 480791.59 287 6 992.06 24

I saw some peaks around 85% CPU, and to have some margins I wouldn't go higher.

[css]

Here are stats from 66 tick 24 slot server, which is full 80% of the time.


  srcds_inout-week.png (Size: 49.28 KB / Downloads: 36)

Take a look at the avg value. It's 514 kbps in and 1230 kbps out. Total is 1744 kbps. That's on average 100% full 19/19 slot 66 tick server. Do the math. (Note: It's bits, not bytes)

janne Wrote:(all players rates forced to 35000,100,100. No source tv)

"Forcing" isn't forcing unless you use Player rate tracking system using Linux's firewall (iptables). Although if everybody use the same "forcing" then the stats obviously match.

[janne]

css Wrote:

janne Wrote:(all players rates forced to 35000,100,100. No source tv)

"Forcing" isn't forcing unless you use Player rate tracking system using Linux's firewall (iptables). Although if everybody use the same "forcing" then the stats obviously match.

I know exactly what you mean.
Even you can't force anyone with, for ex. low fps, to get good rates, just kick or ban them.
I think the built in forcing method is good enough and I don't like your approach to the problem at all.

I still think my measured results are accurate enough for him to make som rough calculations.

(Sorry for the off topic part.)

[ColoDedi]

Just had another look at CPU's and found a lower power chip its a 2.0ghz quad core xeon with 12mb cache its one of the newer chips.
Reckon this would hold up 150 players?