What are the best ways to reduce server power consumption?
I have a Dell Poweredge 860 (2.4ghz QC, 5GB RAM, 2 HDs) and it uses 0.46AMPs when running normally. The problem is at boot up it spikes for a few seconds here and there to about 0.6 and this is a real problem as I only have a 0.5AMP allowance.
Should I underclock it? Remove the CD Drive?
how to go about spec'ing the power consumption of a server. below are the specs of the two servers I'm interested it.
Dual Xenon 2.8Ghz Processors with 2MB cache
1 250Gig SATA Hard Drives
1 500Gig SATA Hard Drive
4 Gigs of Ram
------
Dual Xenon Quad Core 2.66Ghz
1 80Gig Hard Drive
2 320Gig Hard Drives
1 1T Hard Drive
8 Gigs of RAM
I'm building a storage server out of spare parts that are lying around here at the office. Purchasing a brandnew server kinda overshoots it's target, since all basic hardware is available lika a big Chieftec 4U casing with 2x 4drive SATA hotswap bays.
The machine will just be dumb storage for saving our backups.
The plan is to add a decent serverboard, proc and some ram as a baseline. On top of that a 3ware 8-port RAID-controller with 8 1TB harddrives (seagate has some nice ones).
Now the only issue i'm having is powerconsumption. The case has an 460W PSU, but I'm not sure if that's enough.
I need to know what happens when a rack has capped its (lets say) 5A limit. I'm trying to calculate what exactly I can put into a rack and am considering that not all servers are going to be 100% load, as that would be bad performance anyway.
I guess it could do the following (but really don't know):
Cap performance of all servers
Cause system failures
Trip
The rack would have servers, a switch and perhaps a Remote Power Strip and Firewall. I don't know how they would be affected.
I maybe interested in co-locating a server but looking from the power max for a 1U server of around 0.3Amps I'm wondering if it will be easy.
Doing the equation 0.3 * 240 gives me 72watts I believe. (Where 240 is the voltage for the UK). So with this, it doesn't give me much choice considering most processors run around 90watts and that doesn't leave any room for the power from hard drives and memory. Does anyone know what the average power consumption is for hard drives and memory?
I'm not looking for a really power system, but I would like something that if I pull it out of co-location that I can still use it one way or another. So was looking at an AMD Athlon 64 X2 where there is a 34watt version that is seemingly discontinued. Or use an AMD Turion X2, which are again hard to find, with a mini-itx board. Running with the mini-itx board, there is also the Intel Core 2 Duo which I can get quite easily.
Does anyone else co-locate and how to you manage the specification?
My nearest major city is Manchester, so naturally I'm looking for rackspace in the region.
Unless anyone has better suggestions, I'm thinking of going with NorthernColo. They start at £50/month but jump to £70/month if you draw more than 1A of current.
If my basic physics is anything to go by, 0.5A at our 240V means a maximum server power rating of 120 watts.
...are there any dual-core / 2GB RAM box configurations which consume less than 300W thesedays? My own USB mouse for my laptop consumes 50mA.
Otherwise I'm begining to think of their 0.5A pricing as being a bit of a scam, since the 1A price also pays for 2U worth of space.
If I were to build a 2.4 core 2 duo w/ 2gb ram and 4 x 500gb sata and 2 x 160 sata, both in raids. Do you guys think that would be above 2 amps usage? I was thinking maybe the efficient cpu would make up for the large amount of hard drives, all though their IO activity won't be that high. Maybe if someone has a similar setup they could shed me some light .
View 9 Replies View Relatedyou are going to purchase a new rack and the data center offers you a specific amount of power.
you are going to use server's specs like this
Dell PE 1950 III
two processors intel xeon E5410 2.33 GHZ , 24 MB cache, 1333MHz FSB
2x250 sata HDD
8 GB RAM
My question is how many amp's need for one server?
or how much power does one dell PE 1950 III with the previous specs needs?
I've just switched my Apache and PHP config over to linux(ubuntu 14.10).
Apache 2.4.12 is using the mpm-event worker and PHP 5.6 is running in FastCGI/fpm mode.
I'm running MariaDB 10.0.17
I'd like to reduce the time to first byte even more, its currently 80 ms for static files and 110ms for PHP generated pages.
1. 80 ms seems like a lot, since the ping to the server is <15ms
How can I reduce this? Are there some tweaks for the Linux networking stack or Apache that I can apply?
I'm also using SSL, I remember it could be tweaked to improve its performance but I don't remember how..
I currently enabled the following in my main apache config:
Code:
ServerSignature Off
ServerTokens Prod
TraceEnable off
EnableMMAP on
EnableSendfile on
UseCanonicalName On
2. Is there a mod_spdy for Apache 2.4? what about http 2.0?
3. Can I tell PHP to make use of unix sockets to connect to the database server? (I heard unix sockets is faster)
4. I've installed memcached for PHP, can I also use this for Apache?
This morning, apache crashed on the server that I assist a friend of mine (junglecat here) with. We finally managed to get it up and going - for some odd reason, restarting apache/httpd via Cpanel wouldn't get it going, but SSHing and typing via command line would.
I downloaded the apache error log, and it's CHOCK full of this, repeating over and over (and over) again:
DBD::mysql:t bind_columns failed: Statement has no result columns to bind (perhaps you need to successfully call execute first) at email.pl line 55.
DBD::mysql:t fetch failed: fetch() without execute() at email.pl line 56.
Use of uninitialized value in concatenation (.) or string at email.pl line 59.
Use of uninitialized value in concatenation (.) or string at email.pl line 61.
Use of uninitialized value in concatenation (.) or string at email.pl line 76.
Have tried searching and either I'm not using the correct terms, or this doesn't happen that much (I am betting on the former). Any ideas?
This is so unimaginably frustrating. I have a website, fightgame.com, with MediaTemple for about 7 months? I don't know how long it's been, but WAY TOO LONG! If they aren't just flat out down, they are constantly having problems. I don't know why I've stayed with them this long, but I guess I've learned my lesson.
For several weeks my stock vbulletin site has sporadically been down due to database access errors. MediaTemple staff is pretty clueless and just couldn't figure it out. After telling them I had no choice but to leave, they agreed to put me on their MySQL container service for free to see if that solves the problems. They didn't tell me when they were going to move my MySQL database, they just did. That's fine, but shortly afterwards all my tables disappeared. Frantically, I opened tickets, called, etc.
It's been 2 days and NOBODY is responding to me whatsoever!!! I would think MediaTemple is out of business or something, what the heck is going on?? No information about what's going on, no plans, no hints as to when they might look into this, absolutely nothing. Well I shouldn't say absolutely nothing. Yesterday I got a response to 1 of my tickets saying they know there's a problem and have escalated the issue for someone to review...
Unfortunately yesterday was a huge day for my site because of the Antonio Margarito vs Shane Mosley megafight. This is when I get all the traffic and new users. But I was down all day long . REALLY unfortunately, my latest backup is somewhat old . I don't have all my recent posts and users... And I have a bad feeling that MediaTemple doesn't take backups of their servers.
So what do I do? I've been involved in the hosting industry in one way or another for many years being a technologist and have never seen such extremely poor service. I don't understand, doesn't MediaTemple host some big sites like TechCrunch?? I guess they don't care at all about their smaller customers. Should I have my lawyer send them a letter or begin a lawsuit to get my data? I would normally assume a hosting company has backups, but MediaTemple is no ordinary hosting company.. If they do have backups, I so desperately want them. FightGame has grown very well and was very exciting for me, without those backups I don't know what's going to happen.
After not hearing from them (aside from escalating the issue for review) for 2 days, I've threatened posting my support tickets on WHT and now have threatened a letter from my lawyer. Do you guys want to see how ridiculous my support tickets are?
I have a HUGE blog with over 15,000 posts that constantly runs the mysql database like NONE other.
I am just wondering, is there a way to reduce load? I am top 100,000 in Alexa and I get over 2,000 uniques / day. What can I do about this?
My load average seems a bit high considering how few processes are running:
[url]
Is there anything I can do to reduce the load generated by Apache?
You can see the Apache2 server status info here:
[url]
It's serving 20-25 requests per second, which are mostly these tiny requests to tracker.php which issues a database query and returns an image. The relevant httpd.conf settings:
Code:
Timeout 120
KeepAlive Off
MaxKeepAliveRequests 100
KeepAliveTimeout 15
<IfModule prefork.c>
StartServers 8
MinSpareServers 5
MaxSpareServers 20
ServerLimit 512
MaxClients 256
MaxRequestsPerChild 4000
</IfModule>
This is a dual Opteron 2212 server (4 cores total) with 4GB RAM.
I'm going to use my apache server to host files(it works with a php script so we can't use FTP for now...) so I was wondering if I could limit this virtual host's bandwidth per second so the websites on the other virtual hosts still load properly. Any module for apache on windows which can do this?
View 2 Replies View RelatedI'm trying to improve the performance of my website and noticed that https pages take really long to load:
[URL] ....
[URL] ....
It appears this problem is caused by using https, because when I open my site without https it loads pretty quickly.
[URL] ....
From what I can see it appears that the main problem is the time to first byte, so my question is how can I reduce this?
My virtualhost for SSL is set like this:
Quote:
<VirtualHost 46.249.47.12:443>
DocumentRoot "C:Program Files (x86)Apache Software FoundationApache24htdocssite 2"
ServerName www.elitegameservers.net
ServerAlias elitegameservers.net
ErrorLog "logs/xgclan.com-error.log"
[Code] ....
I have replaced the real paths with "PATH" for security reasons. How to reduce the time to first byte in general. I tried setting up memcache to accomplish this but I didn't notice any changes...
This is my mem cache config:
Quote:
<IfModule mod_mem_cache.c>
CacheEnable mem /
MCacheSize 10000000
MCacheMaxObjectCount 100000
MCacheMinObjectSize 1
MCacheMaxObjectSize 262144
</IfModule>
Update: Just ran the openssl speed command [URL] ....
I have been interested in setting up my own server for hosting a website and a small online game I created. My only problem is how good of a server I would need to get.
View 10 Replies View RelatedGiven the choice which is more beneficial. Some of the things I've read say Database servers use up more RAM and others say they use up more CPU power. I'm just not sure.
Also, what HDD should be used for the database server? I have a choice for SATA or SAS.
Some benchmarks are showing SATA as faster I believe, but that SAS is more reliable? Can anyone comment intelligently regarding that as well?
I'm currently running Dell 1750s, 1850s, and 1950s in a colo facility. I am not happy with the 1850s and 1950s power consumption. My 1950s have a single quad core 5310, 2GB memory, dual 15k 73GB drives, dual power suppies and are running at about 1.9 amps with spikes up to 2.4 amps. My applications are disk bound and the servers typically run at a load of .1 to .2.
I'm looking for alternatives to the 1950 that use significantly less power. I need at least 2 Hot Plug SAS drives and would like to have it in 1U. I run 2GB of memory. Dual power supplies would be nice, but are not absolutely necessary. I'd rather not go with a non-hot plug solution, but may have to consider it. I will probably buy 10-15 servers soon and would like them to be identical. I'd prefer buying a name-brand.
I was thinking about changing one of my servers to the desktop centos version instead of pure ssh access. how much added resources will this use?
View 6 Replies View RelatedWould it be bad to run only 24 hard drives on a 900 watt redundant PSU setup provided in a super micro machine? No motherboard, just the hard drives and backplane? I know that if PSU's aren't loaded enough, they can malfunction or shutdown, as well as encounter efficiency issues. Is there a general rule of thumb for this sort of thing? Should I see if supermicro can give me a lower rated PSU or something? I searched around on the net a bit but couldn't find anything solid for this particular question..
View 8 Replies View RelatedI've been having a problem with my webhost (in a shared server), and I don't know if this should be a usual problem or if it is my fault or if it is the host's fault.
The thing is I am getting lots of consecutive petitions for the same files from an IP at a time, which in my logs gets registered like this:
Code:
212.166.183.3 - - [03/Apr/2007:10:07:56 +0200] "GET /downloads/User_manual.pdf HTTP/1.1" 200 1388809 "http://www.mysite.com/desc.html" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.1.4322)" "-"
212.166.183.3 - - [03/Apr/2007:10:07:56 +0200] "GET /downloads/User_manual.pdf HTTP/1.1" 206 66716 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.1.4322)" "-"
212.166.183.3 - - [03/Apr/2007:10:07:58 +0200] "GET /downloads/User_manual.pdf HTTP/1.1" 206 38519 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.1.4322)" "-"
212.166.183.3 - - [03/Apr/2007:10:08:00 +0200] "GET /downloads/User_manual.pdf HTTP/1.1" 206 34072 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.1.4322)" "-"
212.166.183.3 - - [03/Apr/2007:10:08:01 +0200] "GET /downloads/User_manual.pdf HTTP/1.1" 206 28175 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.1.4322)" "-"
212.166.183.3 - - [03/Apr/2007:10:08:02 +0200] "GET /downloads/User_manual.pdf HTTP/1.1" 206 30297 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.1.4322)" "-"
212.166.183.3 - - [03/Apr/2007:10:08:04 +0200] "GET /downloads/User_manual.pdf HTTP/1.1" 206 26516 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.1.4322)" "-"
212.166.183.3 - - [03/Apr/2007:10:08:05 +0200] "GET /downloads/User_manual.pdf HTTP/1.1" 206 40302 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.1.4322)" "-"
The thing is I am getting too much bandwidth consumption, and since I've only been with this host for 3 months now, and I never had similar problems before, I was wondering if this problem is something I can do something about? Or is it the host's config?
Why do these repeat 206 petitions happen?
do dual power supplies use more power than a single supply?
E.g. Say I have a server than uses two amps, powered by a single power supply. Now if I switch to a dual supply (and say each supply has the same efficiency rating as the single), does my server use more power? How much more?
My simple view of this is that it probably does, but maybe not much. The second power supply consumes some power itself, but since its not under load, it doesn't consume much. Therefore, my server with redundant supplies might use 2.1A or 2.2A.
at getting PDUs for the new facility where we plan to move our servers to. Our cabinet will have 2*20a circuits, so we're thinking of getting 2 APC PDUs (1 per circuit). Some offer remote reboot capability.
Our servers (Dell PowerEdge) have dual power supplies, and I assume the best thing is to plug one power supply into each circuit, so things stay up even if a circuit fails.
So, how would this work if we want to reboot? The servers can run on only one power supply, so does that mean we have to click reboot on the web interface of each PDU at exactly the same time?
I've never worked with PDUs before.. just cheap power strips and a call to the datacenter
There are Small Hosting companies and There are Big Hosting companies.
And Then There are Huge Hosting companies.
Who Consumes The Most Resources?
Your Views On It?
my friend has an un-managed VPS plan in kloxo-hypervm mode
yesterday the stats showed in hypervm as the space was over used, so I personally deleted the processed stats which cleared around 20gb space, again today the space is totally consumed, I don't know why its happening and have no clue where are the other locations where the space could have been consumed i.e similar to location of logs or any other specs which could have been consuming space
unknown space consumption - very high no idea how!
its actually an image hosting site and I noted that not many pics and not much space was consumed in the pics uploaded in the last 24hrs but this really puzzles and troubles me as I am currently tracking my friend's VPS, please help me sort this out problem and keep my head high before my friend
I have a vps with 1GB of ram and mysql consume lots of memory I googled about this subject but everywhere mysql performance tuning is described
mysql has 10 threads like below
7338 mysql 0 0.0 4.5 /usr/sbin/mysqld --basedir=/ --datadir=/var/lib/mysql --user=mysql --pid-file=/var/lib/mysql/zeta60.wuhosting.com.pid --skip-external-locking
7339 mysql 0 0.04.5 /usr/sbin/mysqld --basedir=/ --datadir=/var/lib/mysql --user=mysql --pid-file=/var/lib/mysql/zeta60.wuhosting.com.pid --skip-external-locking
7378 mysql 0 0.04.5 /usr/sbin/mysqld --basedir=/ --datadir=/var/lib/mysql --user=mysql --pid-file=/var/lib/mysql/zeta60.wuhosting.com.pid --skip-external-locking
2507 mysql 0 0.04.5 /usr/sbin/mysqld --basedir=/ --datadir=/var/lib/mysql --user=mysql --pid-file=/var/lib/mysql/zeta60.wuhosting.com.pid --skip-external-locking
2517 mysql 0 0.04.5 /usr/sbin/mysqld --basedir=/ --datadir=/var/lib/mysql --user=mysql --pid-file=/var/lib/mysql/zeta60.wuhosting.com.pid --skip-external-locking
2518 mysql 0 0.04.5 /usr/sbin/mysqld --basedir=/ --datadir=/var/lib/mysql --user=mysql --pid-file=/var/lib/mysql/zeta60.wuhosting.com.pid --skip-external-locking
2519 mysql 0 0.04.5 /usr/sbin/mysqld --basedir=/ --datadir=/var/lib/mysql --user=mysql --pid-file=/var/lib/mysql/zeta60.wuhosting.com.pid --skip-external-locking
2520 mysql 0 0.04.5 /usr/sbin/mysqld --basedir=/ --datadir=/var/lib/mysql --user=mysql --pid-file=/var/lib/mysql/zeta60.wuhosting.com.pid --skip-external-locking
7399 mysql 0 0.14.5 /usr/sbin/mysqld --basedir=/ --datadir=/var/lib/mysql --user=mysql --pid-file=/var/lib/mysql/zeta60.wuhosting.com.pid --skip-external-locking
5162 mysql 0 0.04.5 /usr/sbin/mysqld --basedir=/ --datadir=/var/lib/mysql --user=mysql --pid-file=/var/lib/mysql/zeta60.wuhosting.com.pid --skip-external-locking
my.cnf properties are
set-variable = max_connections=90
set-variable = wait_timeout=10
set-variable = connect_timeout=10
set-variable = long_query_time=4
skip-locking
key_buffer = 2M
max_allowed_packet = 1M
table_cache = 4
sort_buffer_size = 64K
read_buffer_size = 256K
read_rnd_buffer_size = 256K
net_buffer_length = 2K
thread_stack = 64K
table_cache = 1M
query_cache_size = 1M
[isamchk]
key_buffer = 8M
sort_buffer_size = 8M
[myisamchk]
key_buffer = 8M
sort_buffer_size = 8M
[mysqlhotcopy]
interactive-timeout
Can anyone share some experiences on this? I'm new to VPS so it's not clear how much RAM cPanel will consume comparing with Webmin after everything is configured for production?
View 10 Replies View RelatedI have a 256 MB VPS. Currently I run BIND to serve a few zones as the primary server and twenty more as the secondary. BIND’s memory footprint is >40 MB which is too much for such low usage. I want to replace BIND with other software to free some memory. Could you please tell me what DNS server has the least memory usage while supporting AXFR? No recursion needed.
What I’ve tried was djbdns. Its zone file format is terrible but perhaps a Web interface such as VegaDNS can help, and TinyDNS does use much less memory than BIND. What else is there?
I just signed my agreement for 2 cabinets at Internap's Atlanta CoLo. For now, I will just have 1 120V 20A circuit in each cabinet.
At my office's server room we currently have 8 circuits I am powering my 2 racks from. I am just curious how many servers you can typically put on a 20A circuit. I am also considering adding two additional power feeds (1 more for each cabinet) to have truly redundant power for my dual power supplies.
I am hoping that 20A is plenty for my needs. I have around 20 HP and Dell systems. Mostly HP DL380s and Dell PE2950s.
Any ideas? I am splitting them in half, only filling up a half of each cabinet for now, as we are growing quickly and I wanted to overguy space so I would not be forced to add another cabinet later and have it end up somewhere else in the DC.
