how to hardware analisys?

Discussion in 'Server Operation' started by matehortua, Oct 12, 2005.

  1. matehortua

    matehortua New Member

    hi everyone

    i want to know what are the hardware prerequisites for a server that is using ISPconfig

    to be more precise how do i make an analisys of the "work" my server does (traffic, ram, HD, swap, proccesor load etc...) whit any new (domain DNS, host web, mail users ) and what would be a good HWconfig.

    sorry about my english, im colombian (thats not an excuse but ;) )
    i hope i make myself clear, if not let me know cause i think its a very important issue to know about hardware capabilities and what is the best we can do whit our hardware

    tanx and bye
  2. falko

    falko Super Moderator

    To see the load of your system and the usage of your RAM and swap, you can run
    If you run
    df -h
    , you can see the usage of your hard disk.
  3. ovis

    ovis HowtoForge Supporter

    I have a Server here for testing puposes Compaq proliant, and here it comes,
    Its 400Mhz and had 9G Hdd 64Mb Ram and it works fine.

    But if your intentions are to host lets say 50 sites you are more concerned with
    Hdd space. A standard install with debian sarge has footprint of about:

    aproxemetly (sorry for my english too i am from holland ;)

    750 MB

    so if you intent to host sites databases and mail add ad mutch hd's in your machine as possible.

    Now the other tree bottlenecks are RAM INTERNET and CPU

    Advice MORE is better altough you can get lot of preformance with linux and old hardware. I hope more experienced users can fill you in on all kinds of info regarding the main bottlenecks of a ISPconfig system

    I hope that i have informed you well :) .

    gr Ovis
  4. themachine

    themachine HowtoForge Supporter

    This isn't all that easy to determine without testing. For most vhoster's I would say that hard drive space is the biggest requirement. Adding RAM and upgrading CPU is quite easy down the road. I see many customers hosting 1500+ sites on less than: 80G IDE hard drive, 1024MB RAM, AMD XP3000+ CPU. All standard desktop hardware which is fairly cheap to get a config like that. Personally, if your doing this as a business I would put your money into SCSI RAID for the hard disks, or if not... please maintain a solid backup solution. Hosters don't realize that IDE drives don't last long when your maxing them out all day, every day. BACKUP!!!!! CONSISTENT AND TESTED BACKUPS!!!! ;)


    SAR: Sar is an excellent utility for everything. It runs cron jobs at 10 minute intervals to give you an idea what load was on the server at that time, and catalogs them by day.

    # sar
    00:00:01  cpu %usr  %sys  %nice  %idle									_cpu_
    00:10:01  all	2	 0	  0	 98
    00:20:01  all	2	 0	  0	 98
    00:30:01  all	0	 0	  0	100
    00:40:01  all	2	 0	  0	 98
    00:50:01  all	2	 0	  0	 98
    01:00:01  all	0	 0	  0	100
    # sar -p
    00:00:01   pagein/s pageout/s	 swapin/s swapout/s		 fork/s	  _page_
    00:10:01	   4.47	 40.12		 0.00	  0.00		   0.39
    00:20:01	   0.01	 35.91		 0.00	  0.00		   0.29
    00:30:01	   0.02	 32.86		 0.00	  0.00		   0.22
    00:40:01	   1.66	 43.78		 0.00	  0.00		   0.54
    00:50:01	   2.71	 28.13		 0.04	  0.00		   0.27
    01:00:01	   0.00	 33.20		 0.00	  0.00		   0.23
    # sar -r
    00:00:01  memtot memfree buffers   cached  slabmem	  swptot swpfree  _mem_
    00:10:01	503M	 15M	 47M	 154M	   0M	   1027M	953M
    00:20:01	503M	 19M	 47M	 153M	   0M	   1027M	953M
    00:30:01	503M	 17M	 48M	 154M	   0M	   1027M	953M
    00:40:01	503M	 10M	 49M	 156M	   0M	   1027M	953M
    00:50:01	503M	 10M	 50M	 159M	   0M	   1027M	953M
    01:00:01	503M	  7M	 51M	 160M	   0M	   1027M	953M
    # sar --help
    sar: invalid option -- -
    usage: sar [-flags] t [n]			  or
    	   sar [-flags] [-s hh:mm] [-e hh:mm] [-i sec] [-n day# | -f file]
    		-A	  all flags
    		-S	  time-stamp for every output-line
    		-u	  cpu (default flag)
    		-P	  process load
    		-d	  disk
    		-D	  disk-partition
    		-r	  memory & swap
    		-p	  paging & swapping
    		-I	  interrupts
    		-v	  kernel-resources
    		-y	  tty activity
    		-l	  net-interf (general)
    		-L	  net-interf (errors)
    		-w	  ip   v4	(general)
    		-W	  ip   v4	(errors)
    		-t	  tcp  v4	(general)
    		-T	  tcp  v4	(errors)
    		-U	  udp  v4
    		-m	  icmp v4	(general)
    		-M	  icmp v4	(per type)
    		-g	  ip   v6	(general)
    		-G	  ip   v6	(errors)
    		-j	  tcp  v6	(general)
    		-h	  udp  v6
    		-k	  icmp v6	(general)
    		-K	  icmp v6	(per type)
    		-N	  nfs		(general)
    		-E	  nfs		(errors)
    		-V	  nfs		(server)
    		-R	  nfs-rpc	(%calls)
    You can even specify the DAY/HOUR/MINUTE.  So if your server crashed at say 12:45 on the 10th of this month, you could:
    # sar -s 12:30 -n 10
    Linux  2.4.27  #1 Wed Apr 6 10:05:14 CDT 2005  i686  10/10/2005
    12:30:01  cpu %usr  %sys  %nice  %idle									_cpu_
    12:40:01  all	2	 0	  0	 98
    12:50:02  all	2	 0	  0	 98
    13:00:01  all	0	 0	  0	100

    IPTRAF: Iptraf is a great utility for network traffic monitoring. This will give you an idea of how much bandwidth you are taking in/putting out in a kb/s format.
  5. falko

    falko Super Moderator

Share This Page