I will second using RAM drives on Linux for running a Minecraft server. Ram drives will handily outperform any SSD setup, and 16GB DDR3 has hit ~$68 on sale recently. I ran a Tekkit server last year on a single 500GB WD Black drive and had constant issues with chunk loading and lag. Put the server in ram and most problems went away. Do keep in mind you need a UPS and scripts auto-backing the server up every couple of minutes to prevent data loss from power interruption. Something to keep in mind with SSD drives is that the drives have a certain write endurance per cell on them. While I don't have numbers on the average writes per second for a Minecraft server, I assume they will be fragmented and frequent with the way the server writes out data. If you have too many writes you will "wear out" an SSD far before it's MTBF.
Server hardware discussion. Any server Owner Must Read !
- Thread starter Greedseed
- Start date
-
The FTB Forum is now read-only, and is here as an archive. To participate in our community discussions, please join our Discord! https://ftb.team/discord
you will NOT run out a SSD on minecraft.
The way they create MLC chips these days made them very robust and they now last up to 10.000 cycles, unlike the 5000 stated before. And even if you think that is way to short. Get some SLC chips, these will last even 10 times longer ! so no you will NOT crap out a SSD on minecraft, and especially not of trim is enabled. Again the Ramdisk is a overkill opion that you will NEVER use to its capabilities. Even the SSD will never be used to its full potential. And i bet you i will get your 16GB or ram disk filled up under 1 month. With normal server usage !
TLC: 5000 Cycles - Cheap cells that are not used as much anymore, mostly build for very cheap and smaller drives but never used in datacenters. will last about 3 years easily with trim.
MLC: 10.000 Cycles - Medium range cells, these are the most used cells these days and double the TLC's lifespan. Will outlast most of the drives out there.
SLC: 100.000 Cycles - This are the newest and datacenter graded cells, less storage but a very high cycle rate. Virtually unkillable.
The way they create MLC chips these days made them very robust and they now last up to 10.000 cycles, unlike the 5000 stated before. And even if you think that is way to short. Get some SLC chips, these will last even 10 times longer ! so no you will NOT crap out a SSD on minecraft, and especially not of trim is enabled. Again the Ramdisk is a overkill opion that you will NEVER use to its capabilities. Even the SSD will never be used to its full potential. And i bet you i will get your 16GB or ram disk filled up under 1 month. With normal server usage !
TLC: 5000 Cycles - Cheap cells that are not used as much anymore, mostly build for very cheap and smaller drives but never used in datacenters. will last about 3 years easily with trim.
MLC: 10.000 Cycles - Medium range cells, these are the most used cells these days and double the TLC's lifespan. Will outlast most of the drives out there.
SLC: 100.000 Cycles - This are the newest and datacenter graded cells, less storage but a very high cycle rate. Virtually unkillable.
Ram drives will outperform SSD, but SSD's are fast enough for minecraft that Ram drives will not provide any additional benefit. Remember, you are still limited by your upload speed. Chunks still have to be sent to the player, so even if the server loads them instantly, they will load just as quick for the players.
Huh?
http://www.anandtech.com/show/6462/intel-explains-20nm-nand-endurance-concerns-on-the-ssd-335
http://www.anandtech.com/show/6459/samsung-ssd-840-testing-the-endurance-of-tlc-nand
http://www.anandtech.com/show/5067/understanding-tlc-nand/3
As memory chips get smaller and denser, the lifespan of the chips is decreasing. This is only offset by the fact that drive sizes are increasing, so overall it's a net wash.
That said, you have to write about 240 GB/day for 5 years to wear a drive out. A Minecraft server isn't going to do that, even if it does write a bunch of small writes. In reality, Minecraft only writes to disk when a chunk is loaded/de-loaded. For chunks near chunk-loaders or the spawn, they stay in memory indefinitely.
I'm only stating a RAMDISK is a solution, because if you're on a Shared VPS where IOPS isn't guaranteed, a Ramdisk can be much better for a smaller server. Yes, if you have your own Managed Server, go all out on RAID SSDs with 0+1 or whatever.
And yes, if you use a ramdisk make a back-up script!
I think tick threading mod takes care of this problem and makes 16 core cpu used to the max. Just the thought.
If hosting on a VPS a lot of VPS's are promissing secure SSD storage.
I personally run a vps with ramnode, and I get great great performance from their ssd's, and they run RAID 10 (check them out they are awesome and cheap
)
So unless you run a big server with lots of players, and you are for some reason unable to get a proper SSD setup, I really don't see any idea in going for a ramdisk.
I personally run a vps with ramnode, and I get great great performance from their ssd's, and they run RAID 10 (check them out they are awesome and cheap
)So unless you run a big server with lots of players, and you are for some reason unable to get a proper SSD setup, I really don't see any idea in going for a ramdisk.
Figured I'd pipe in here since this is what I've done for a living for the past 20 years roughly (since 1994) (Unix/Linux Systems and Network Administration), everything from webhosting, to isps, to game servers, to corporate private networks, wans, etc. I really don't recommend running a server yourself unless you have some idea of how to function at the commandline in linux and actual server management in windows and at least some background in security.
First lets establish some correct definitions:
VPS/VDS Virtual Private Server/Virtual Dedicated Server (Hardware: Vendor) - Used fairly interchangeably now. A virtual server dedicated to you that has an allocated set of resources reserved for your machines operation. These typically run in a cluster of vmware/xen/hyperv machines that have hundreds of gigs of ram each, store the virutal machine files on a SAN, and have upwards of 8+ cores per processor and usually between 2-8 processors per server. Each may run just a few VDS/VPS or they may run hundreds. Unless overloaded fairly massively, this doesn't result in a performance decreace for the various hosts (often they still perform better than a dedicated server due to underlying higher end storage systems and such... but if overloaded they're abysmal performance wise.) They often can do automatic failover, snapshots of running systems for backups, and live migrations between various hosts in the cluster to help maintain nearly 100% uptime. (We usually aim for between 4 and 5 9s availability where i've worked ) They carve out the specifics of your system from the resources on a host and if need be it should be able to be freely moved to another host if something is amiss on the host it started on and often times will be automatically migrated to other hosts for performance reasons. You should remain blissfully unaware of this.
Dedicated Server - (Hardware: Vendor) The vendor owns the hardware you're using. You can typically install whatever you need on the system to accomplish your goals. You want a RAC.
Co-located Server - (Hardware: You) You provide a given piece of hardware with a specific specified foot print and they provide you rack space and bandwidth. Typically 1-4 RU is what they provide with some quantity of bandwidth. Make sure you have a RAC card of some type in the system and that it's accessible so you can get a remote console if the event of a networking failure. Whatever fails, you will be responsible for replacing. You will likely also be entirely responsible for backups and operating system updates. Some facilities have non-allowed content and services (many exclude for instance irc daemons since irc has a tendency to get denial of service, almost all exclude anything illegal, etc.)
Managed Server - (Software: Vendor) A server managed by vendor, can be any of the 3 types above, typically this means they will provide at least some level of software/operating system installation and on-going maintenance. Usually they break managed services into various levels of increasing price for increasing levels of management. At the high end they might help you optimize your database and webservers, at the low end they might install security updates once a month. Often times a managed server will come with a control panel as well if you're not familiar with server management a managed server with a control panel is a good way to go.
Vendor/Host - The people providing you the services.
I'm primarily going to talk about hosting on linux, because after 20 years of doing both I wouldn't consider hosting an application on windows that absolutely didn't require it. Minecraft doesn't require it.
Processor - 2 cores/processors tends to be a bit light, 4-8 seems to land pretty squarely in the minecraft realm. This lets you dedicate several cores to minecraft (and mods) and still have a core available for additional processes like databases or webservices you're providing (forums, teamspeak, etc.) Hyperthreaded cores are ok for reducing cpu time contention, but don't count them as part of the available cores. If i felt like counting them at all, i'd count them as .25 cores each. Cache matters, 4096kb+... 8192kb is better. Raw Ghz matters. Higher is better... but it's a secondary consideration until you max out ram.
Memory - RAM, I figure 2g + 128mb/concurrent person roughly until I hit 8g for "heavily modded" servers (my own pack has ~110 mods and runs well on 8g with upwards of 20-30 people concurrently). After that you gain more performance by dedicating ram to mysql for block logging, ramdisks, increasing cpu, or spending time optimizing your java performance. If I had a server with say 128g of ram, I'd likely dedicate 16g to java (minecraft), 8g to the database (for most minecraft purposes this is usually more than adaquete), and then allocate the rest to a ram disk.
Network - Very important. Verify you can get enough bandwidth from a host before you sign up with them. I figure .5Mb/s *BOTH DIRECTIONS* per concurrent user... I've seen it spike beyond that though with certain mods, but never on vanilla and even with mods ive never seen it sustain at >.5Mb/s per user. Best case is test first as always. Also be aware of bandwidth charges, you will get a certain amount as part of your package, but if you exceed that you will be charged an absurd amount of money per G you exceed it. If you THINK you might exceed it, call the host and ask for your package to be upgraded to a higher bandwidth quantity. It will save you money. You want at least 100mbit, don't take 10mbit these days (it indicates a poorly designed network and misuse of equipment/reuse of old equipment.) Gigabit is better (and pretty frequently offered.) Keep in mind these are local lan speeds, and what matters just as much is the hosts connectivity. Watch your hosts ping times for a week and see how much packet loss you're seeing, see what the average ping times are, see how many hops they are from you and from the backbone via traceroute. Lower on ping is better. Over 100 is unacceptable.
Storage - First, understand this: Comparing any solution except a ram disk as a single drive option is idiotic. You're virtually begging for data loss. On a vps/vds you absolutely should be on a raid no matter what since it should be stored on a SAN (even if you're unaware or can't see it due to the virtualization.) If not-- find a new vds/vps host. On a dedicated or co-located machine you should have a minimum of 2 drives in a mirrored configuration (raid 1)or 3-4 drives in raid 5. Increasing the number of drives increases the spindles (for non-ssd and stripes for ssd [with caveat]) which will improve performance usually. At some point you're better changing to a different type of disk/interface. Interface wise you generally follow this route: SATA II->SATA III->SAS/SCSI->PCIe->Direct (ram). Drive types you generally go: 5400->7200->10000-->SAS/SCSI(10000)->SAS/SCSI(15000)->SSD SATA->SSD SAS->SSD PCIe->Ramdisk. Now that all said, lets discuss raid and ssd drives. If you're using raid on a ssd you lose (or last i researched it this was true, ~8mo ago) the ability for the ssd's to use the TRIM command which is the primary mechanism ssd's prolong their life with. I've heard claims that the newer garbage collection makes up for some of the loss but I've seen no evidence of such yet and after this long and seeing the number of ssd failures in the field that I have... I'm a real sceptic. Meaning the drive will degrade what most people would consider "massively fast" (couple years) vs normal ssd life of 5+ years... that's fine as long as you plan for it though, planned failure is something you do with ram disks as well... If you're planning on using a ram disk you need to make some additional considerations, primarily what I refer to as "Planned Failure", without a doubt at some point... you will have something happen that will result in the server reseting unexpectedly. If you're using a ram disk and aren't backing up very frequently (5m-15m) you are going to lose all the data since your last backup. As long as you're doing frequent backups and have planned well for the eventual, you really can't find better performance than a ram drive. Putting all the world directory on the ram disk will only require a small ram disk and offers a fairly significant performance boost. If you're planning on including the entire minecraft directory and have dynmap, I hope you allocated plenty of space. Our world directory is about 1g, our minecraft (mostly due to dynmap) is 7.2g and it's not setup to be massively HD and most of the mystcraft worlds don't have significant renders. So long as you plan for and properly implement these things you will have no problems.
First lets establish some correct definitions:
VPS/VDS Virtual Private Server/Virtual Dedicated Server (Hardware: Vendor) - Used fairly interchangeably now. A virtual server dedicated to you that has an allocated set of resources reserved for your machines operation. These typically run in a cluster of vmware/xen/hyperv machines that have hundreds of gigs of ram each, store the virutal machine files on a SAN, and have upwards of 8+ cores per processor and usually between 2-8 processors per server. Each may run just a few VDS/VPS or they may run hundreds. Unless overloaded fairly massively, this doesn't result in a performance decreace for the various hosts (often they still perform better than a dedicated server due to underlying higher end storage systems and such... but if overloaded they're abysmal performance wise.) They often can do automatic failover, snapshots of running systems for backups, and live migrations between various hosts in the cluster to help maintain nearly 100% uptime. (We usually aim for between 4 and 5 9s availability where i've worked ) They carve out the specifics of your system from the resources on a host and if need be it should be able to be freely moved to another host if something is amiss on the host it started on and often times will be automatically migrated to other hosts for performance reasons. You should remain blissfully unaware of this.
Dedicated Server - (Hardware: Vendor) The vendor owns the hardware you're using. You can typically install whatever you need on the system to accomplish your goals. You want a RAC.
Co-located Server - (Hardware: You) You provide a given piece of hardware with a specific specified foot print and they provide you rack space and bandwidth. Typically 1-4 RU is what they provide with some quantity of bandwidth. Make sure you have a RAC card of some type in the system and that it's accessible so you can get a remote console if the event of a networking failure. Whatever fails, you will be responsible for replacing. You will likely also be entirely responsible for backups and operating system updates. Some facilities have non-allowed content and services (many exclude for instance irc daemons since irc has a tendency to get denial of service, almost all exclude anything illegal, etc.)
Managed Server - (Software: Vendor) A server managed by vendor, can be any of the 3 types above, typically this means they will provide at least some level of software/operating system installation and on-going maintenance. Usually they break managed services into various levels of increasing price for increasing levels of management. At the high end they might help you optimize your database and webservers, at the low end they might install security updates once a month. Often times a managed server will come with a control panel as well if you're not familiar with server management a managed server with a control panel is a good way to go.
Vendor/Host - The people providing you the services.
I'm primarily going to talk about hosting on linux, because after 20 years of doing both I wouldn't consider hosting an application on windows that absolutely didn't require it. Minecraft doesn't require it.
Processor - 2 cores/processors tends to be a bit light, 4-8 seems to land pretty squarely in the minecraft realm. This lets you dedicate several cores to minecraft (and mods) and still have a core available for additional processes like databases or webservices you're providing (forums, teamspeak, etc.) Hyperthreaded cores are ok for reducing cpu time contention, but don't count them as part of the available cores. If i felt like counting them at all, i'd count them as .25 cores each. Cache matters, 4096kb+... 8192kb is better. Raw Ghz matters. Higher is better... but it's a secondary consideration until you max out ram.
Memory - RAM, I figure 2g + 128mb/concurrent person roughly until I hit 8g for "heavily modded" servers (my own pack has ~110 mods and runs well on 8g with upwards of 20-30 people concurrently). After that you gain more performance by dedicating ram to mysql for block logging, ramdisks, increasing cpu, or spending time optimizing your java performance. If I had a server with say 128g of ram, I'd likely dedicate 16g to java (minecraft), 8g to the database (for most minecraft purposes this is usually more than adaquete), and then allocate the rest to a ram disk.
Network - Very important. Verify you can get enough bandwidth from a host before you sign up with them. I figure .5Mb/s *BOTH DIRECTIONS* per concurrent user... I've seen it spike beyond that though with certain mods, but never on vanilla and even with mods ive never seen it sustain at >.5Mb/s per user. Best case is test first as always. Also be aware of bandwidth charges, you will get a certain amount as part of your package, but if you exceed that you will be charged an absurd amount of money per G you exceed it. If you THINK you might exceed it, call the host and ask for your package to be upgraded to a higher bandwidth quantity. It will save you money. You want at least 100mbit, don't take 10mbit these days (it indicates a poorly designed network and misuse of equipment/reuse of old equipment.) Gigabit is better (and pretty frequently offered.) Keep in mind these are local lan speeds, and what matters just as much is the hosts connectivity. Watch your hosts ping times for a week and see how much packet loss you're seeing, see what the average ping times are, see how many hops they are from you and from the backbone via traceroute. Lower on ping is better. Over 100 is unacceptable.
Storage - First, understand this: Comparing any solution except a ram disk as a single drive option is idiotic. You're virtually begging for data loss. On a vps/vds you absolutely should be on a raid no matter what since it should be stored on a SAN (even if you're unaware or can't see it due to the virtualization.) If not-- find a new vds/vps host. On a dedicated or co-located machine you should have a minimum of 2 drives in a mirrored configuration (raid 1)or 3-4 drives in raid 5. Increasing the number of drives increases the spindles (for non-ssd and stripes for ssd [with caveat]) which will improve performance usually. At some point you're better changing to a different type of disk/interface. Interface wise you generally follow this route: SATA II->SATA III->SAS/SCSI->PCIe->Direct (ram). Drive types you generally go: 5400->7200->10000-->SAS/SCSI(10000)->SAS/SCSI(15000)->SSD SATA->SSD SAS->SSD PCIe->Ramdisk. Now that all said, lets discuss raid and ssd drives. If you're using raid on a ssd you lose (or last i researched it this was true, ~8mo ago) the ability for the ssd's to use the TRIM command which is the primary mechanism ssd's prolong their life with. I've heard claims that the newer garbage collection makes up for some of the loss but I've seen no evidence of such yet and after this long and seeing the number of ssd failures in the field that I have... I'm a real sceptic. Meaning the drive will degrade what most people would consider "massively fast" (couple years) vs normal ssd life of 5+ years... that's fine as long as you plan for it though, planned failure is something you do with ram disks as well... If you're planning on using a ram disk you need to make some additional considerations, primarily what I refer to as "Planned Failure", without a doubt at some point... you will have something happen that will result in the server reseting unexpectedly. If you're using a ram disk and aren't backing up very frequently (5m-15m) you are going to lose all the data since your last backup. As long as you're doing frequent backups and have planned well for the eventual, you really can't find better performance than a ram drive. Putting all the world directory on the ram disk will only require a small ram disk and offers a fairly significant performance boost. If you're planning on including the entire minecraft directory and have dynmap, I hope you allocated plenty of space. Our world directory is about 1g, our minecraft (mostly due to dynmap) is 7.2g and it's not setup to be massively HD and most of the mystcraft worlds don't have significant renders. So long as you plan for and properly implement these things you will have no problems.
(apparently i wrote to much novel for a post lol)
How to check your allocated resources on a linux host:
cat /proc/cpuinfo <- shows your cpu info
cat /proc/meminfo <- shows mem totals and usage info
fdisk -l <- shows raw disk information
df -h <- shows mounted disk information
iptraf (iftop, nethogs, ntop) <- shows current network usage
iostat <- shows io stats for storage system
sar <- collects system statistics so you can get an idea of what was going on at 4am last night that made people cry.
bonnie++ <- does disk performance tests (massive load, don't do on a production system.) You can find results to compare to online.
The last thing I want to touch on is the thing I as a consultant spent more time "fixing" than anything else, backups. You NEED good backups. You NEED to verify them. This is the thing people spend the least amount of time on and live under the delusion of "oh i have a ssd they don't have moving parts and won't fail." "oh I've got raid that can't fail." "oh i'm on a san that can tolerate 7 drive failures before a data loss, I can't lose data." You can and will encounter situations that will result in data lose with all of those attitudes. I have dug more information out crashed systems and corrupted databases and flat files than I care to remember. It's always a miserable disappointment in the end when that ONE critical piece of data isn't recoverable even though most of the rest is. Backup. Backup. Backup. If you don't know how to correctly backup your data wiht a consistent state (eg: no half written data) then get a control panel or application tha will do it for you and get the files off the system on a regular (daily) basis. At some point in your server management time it will save you more heart ache than you can imagine. I promise you that. As far as scheme to keep, I like last 7 days, last 5 sundays, and the last backup of each month (nothing like finding a corrupt file when everything else is valid and it's a 9 days old and not having a valid backup to fall back on.)
Anyways, I've written a small novel now. I'll shutup, there's so much more you should consider though really-- firewalls (iptables), updates, remote management, security practices, control panels, etc... for anyone that cares, anymore I usually run smaller (<15) servers on EC2 using the amazon version of linux with m2.xlarge spot instances with hourly backups offsite of all the important info... I can redeploy our server in ~30m from "request a spot instance" till people can login. It's been down twice in the last 6mo, once for a redeploy (moved from west to east coast) and once for exceeding my price point. Runs about $40/mo for 17g of ram, 2 cores, 80g disk (ebs), run a small ramdisk, forum, teamspeak, mc@8g. Flies. Price makes it pretty appealing.
Good luck!
How to check your allocated resources on a linux host:
cat /proc/cpuinfo <- shows your cpu info
cat /proc/meminfo <- shows mem totals and usage info
fdisk -l <- shows raw disk information
df -h <- shows mounted disk information
iptraf (iftop, nethogs, ntop) <- shows current network usage
iostat <- shows io stats for storage system
sar <- collects system statistics so you can get an idea of what was going on at 4am last night that made people cry.
bonnie++ <- does disk performance tests (massive load, don't do on a production system.) You can find results to compare to online.
The last thing I want to touch on is the thing I as a consultant spent more time "fixing" than anything else, backups. You NEED good backups. You NEED to verify them. This is the thing people spend the least amount of time on and live under the delusion of "oh i have a ssd they don't have moving parts and won't fail." "oh I've got raid that can't fail." "oh i'm on a san that can tolerate 7 drive failures before a data loss, I can't lose data." You can and will encounter situations that will result in data lose with all of those attitudes. I have dug more information out crashed systems and corrupted databases and flat files than I care to remember. It's always a miserable disappointment in the end when that ONE critical piece of data isn't recoverable even though most of the rest is. Backup. Backup. Backup. If you don't know how to correctly backup your data wiht a consistent state (eg: no half written data) then get a control panel or application tha will do it for you and get the files off the system on a regular (daily) basis. At some point in your server management time it will save you more heart ache than you can imagine. I promise you that. As far as scheme to keep, I like last 7 days, last 5 sundays, and the last backup of each month (nothing like finding a corrupt file when everything else is valid and it's a 9 days old and not having a valid backup to fall back on.)
Anyways, I've written a small novel now. I'll shutup, there's so much more you should consider though really-- firewalls (iptables), updates, remote management, security practices, control panels, etc... for anyone that cares, anymore I usually run smaller (<15) servers on EC2 using the amazon version of linux with m2.xlarge spot instances with hourly backups offsite of all the important info... I can redeploy our server in ~30m from "request a spot instance" till people can login. It's been down twice in the last 6mo, once for a redeploy (moved from west to east coast) and once for exceeding my price point. Runs about $40/mo for 17g of ram, 2 cores, 80g disk (ebs), run a small ramdisk, forum, teamspeak, mc@8g. Flies. Price makes it pretty appealing.
Good luck!
Hey,
I am running a ftb ultimate server but i am not sure iff my server is powerfull enough to run it.
These are the hardware details of my server:
OS: Windows Server 2008 r2 standard.
Type System: 64bit.
RAM: 8gb 1333mhz
CPU: Intel core duo E4600 2.40ghz
And the computer where i play on has the following hardware:
OS: Windows 8 Pro.
Type System: 64bit.
RAM: 8gb 1333mhz
CPU: Intel Core i7 2670QM 2.2 -3.1 ghz
GPU: Nvidia Geforce GT540M 1gb
Please tell me what the problem is
Greetings,
Don
I am running a ftb ultimate server but i am not sure iff my server is powerfull enough to run it.
These are the hardware details of my server:
OS: Windows Server 2008 r2 standard.
Type System: 64bit.
RAM: 8gb 1333mhz
CPU: Intel core duo E4600 2.40ghz
And the computer where i play on has the following hardware:
OS: Windows 8 Pro.
Type System: 64bit.
RAM: 8gb 1333mhz
CPU: Intel Core i7 2670QM 2.2 -3.1 ghz
GPU: Nvidia Geforce GT540M 1gb
Please tell me what the problem is
Greetings,
Don
I'm thinking about setting up a small server for me and my friends.
I want to do this based on Linux and this is the hardware I come up with:
CPU: AMD Athlon II X2 240 Tray
Mobo: Asrock N68-VS3 FX
HDD: Hitachi HTS545032A7E380, 320GB SATA-300
Ram::Crucial CT2KIT25664BA1067
Case: HKC 3011GS (400W PSU included)
The total price of this small server would be 150 euro's.
Would this be enough for a maximum of 10 players?
It will be a FTB Lite server (with additional add-ons).
This server is now running on my game system on which I'm also playing FTB.
I want to do this based on Linux and this is the hardware I come up with:
CPU: AMD Athlon II X2 240 Tray
Mobo: Asrock N68-VS3 FX
HDD: Hitachi HTS545032A7E380, 320GB SATA-300
Ram::Crucial CT2KIT25664BA1067
Case: HKC 3011GS (400W PSU included)
The total price of this small server would be 150 euro's.
Would this be enough for a maximum of 10 players?
It will be a FTB Lite server (with additional add-ons).
This server is now running on my game system on which I'm also playing FTB.
Anyone tried hosting on a HP Proliant Microserver N40L? Can get a hold of one rather cheap (100USD) from a local reseller right now (Diskless, 4GB RAM)
I haven't personally tried it, but the processor in that thing is weak as hell. Would suggest renting a VPS or dedicated box instead. You could also build a small server for under $500 and if your home internet is fast enough you could run it from there.
I'm looking at a system with
(1) HP ProLiant DL360 G5 Rack CTO Chassis
(2) Quad-Core Intel Xeon Processor E5450 (3.0 GHz, 80 Watts, 1333 FSB)
(4) 4GB FBD PC2-5300 2 x 2 GB Kit, 16GB total
(1) HP Smart Array P400i/256MB Controller
(2) 700W HP redundant power supplies with IEC cord only
(1) DVD-ROM Combo Option Kit
(2) HP 36GB 3G SAS 10K SFF SP HDD
(1) Rail Kit
on ebay. Thoughts? My server has about 15 mods including mo creatures and industrial craft running on MC 1.5.2 with MCPC+ as the backend. We have 8 actives, looking to expand.
(1) HP ProLiant DL360 G5 Rack CTO Chassis
(2) Quad-Core Intel Xeon Processor E5450 (3.0 GHz, 80 Watts, 1333 FSB)
(4) 4GB FBD PC2-5300 2 x 2 GB Kit, 16GB total
(1) HP Smart Array P400i/256MB Controller
(2) 700W HP redundant power supplies with IEC cord only
(1) DVD-ROM Combo Option Kit
(2) HP 36GB 3G SAS 10K SFF SP HDD
(1) Rail Kit
on ebay. Thoughts? My server has about 15 mods including mo creatures and industrial craft running on MC 1.5.2 with MCPC+ as the backend. We have 8 actives, looking to expand.
Assuming for the hard drive you meant 360GB everything looks good. 36GB would be quite small, not nearly enough for a modern operating system and a large server map.
The processor is a bit old, so don't tax it to much, it will have limitations. I would guess about 20 players max with that setup, depending on your internet speed.
The processor is a bit old, so don't tax it to much, it will have limitations. I would guess about 20 players max with that setup, depending on your internet speed.
It's 36. I'm just running a lightweight linux server on it, and backups are offloaded to another storage array. Players are helping to purchase something new, so I'm considering going newer, issue is anything Sandy bridge is going to add a good chunk to the price. Haswell is quite obviously out of the question as that's just hitting the market.
My personal choice would be a dual core chip with hyperthreading or a real quad core, a 120gb solid state drive, and 16 gigs of memory. Assuming your running this headless, you don't need a screen, keyboard, mouse. You could get a desktop motherboard, an Ivy Bridge processor, a 400 watt power supply, and a cheap case and the rest is completely optional. When I have some spare time I can show you a setup that fits your budget if you want.