Thursday, September 17, 2009
5:42 PM

BootChart, Performance, Benchmarks



BootChart

Below is a chart showing lots of information of the bootup process in Ubuntu. The main thing here it is showing is performance and the amount of time it took to boot. For me, it was 26 seconds. I believe I can do better, but I have many things enabled, and I haven't attempted overclocking again. Yet, my system is very responsive and fast when up. To get bootchart...


sudo apt-get install bootchart
 










Performance
One of the biggest things I do to increase my performance is run hard drives in Striped Raid Level 0. This Raid Level gives the best performance of them all, but many complain because of the risks associated with it. As opposed to a type of mirrored RAID (Rendundant Independent Array of Disks), if one drive goes out, you lose all data with an exception. I have automated backups performed every night using cron and my own rsync command with lots of options that puts backups for example of everything in /home to my 1.5 TB TeraByte (1024 Gigabytes = 1 TeraByte) drive. This has been useful even though I have not had a drive in Raid 0 fail me because when a file is mistakenly deleted, I can go to the backups, and restore it immediately. Rsync is the best for backups. It can not only do backups to a remote server, but in my case, locally to another drive used for storage and backups that's not in a Raid configuration. I plan to create a screencast of how to setup Raid. I've done it in both Ubuntu, Fedora, and CentOS, but it can be setup without any distro's installer prior to installing the distro just as partitioning can. Other things I do to increase performance are disable unused services, support for hardware I don't have (see previous blog post), use Firefox 3.5 with cache set to use my RAM (aka memory, Random Access Memory, not to be confused with hard drive memory), occasionally use a lightweight Windows Manager such as Xmonad (Great Tile Manager) and JWM (The one used in Puppy Linux) which can all be installed through the package manager. Overclocking is okay if you are willing to take the risks, know what you are doing, and some hardware is MUCH better for it than others.


Setting the Firefox cache to use memory.
First, let me say that memory is volatile meaning that when the memory loses power (a computer shuts down), nothing is saved in it as opposed to hard drives, so this can also be though of as a form of security as once you shut down or reboot, the cache is gone, and a person then can't look through your cache to see where you have been on the interweb. Here are the instructions for setting this up:
  1.  In Firefox, open a new tab (Control + t does that) and switch to it.
  2.  In the address bar, type "about:config" without quotes, and accept the message to continue.
  3.  Right click in a white area and then left click "New>String"
  4.  Put in without quotes as the string "browser.cache.disk.parent_directory"
  5.  Next you are asked for the value. Put in without quotes "/dev/shm" (This is the device node Linux uses for your memory.)
  6.  Now close that tab (Control + w) and Restart Firefox.


Benchmarks
 There are several separate tools to use for benchmarking depending on which hardware you want to benchmark. Here are a few of my favorites.

Video card: I use glxgears in terminal and wait for it to display some information. Earlier I got over 8k FPS (Frames Per Second), not Frames Per 5 Seconds. I thought I had seen higher before, not sure what's going on.

Hard Disks:  


$ sudo hdparm -Tt /dev/md0

/dev/md0:
 Timing cached reads:   2106 MB in  2.00 seconds = 1053.39 MB/sec
 Timing buffered disk reads:  316 MB in  3.03 seconds = 104.43 MB/sec 

Even though I am running my system on top of Raid 0, it is likely that your speeds are close to mine if you have purchased modern Sata2 Drives. The reason is because the cache for those modern drives is usually 32 mb, and mine are about a year old and have a cache of only 8mb, 16mb cache for the 1.5tb drive. I have not figured out yet how to measure write speeds in a nice format like the above. If you know how, leave a comment. That's what Raid 0 is supposed to be good at when compared to Raid 1 (mirrored, not striped). Your metal oxide platters likely spin at the same standard 7200 RPMs unless you are running a laptop. Then I believe it is 5400 RPMs last time I checked.


Checking overall, mostly a combination of CPU, RAM, and BUS SPEEDS:
Install p7zip-full.
Open in your web browser the following.
file:///usr/share/doc/p7zip-full/DOCS/MANUAL/commands/bench.htm




$ 7z b -mmt2 -md26 2

7-Zip  4.58 beta  Copyright (c) 1999-2008 Igor Pavlov  2008-05-05
p7zip Version 4.58 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,2 CPUs)

RAM size:    2007 MB,  # CPU hardware threads:   2
RAM usage:    777 MB,  # Benchmark threads:      2

Dict        Compressing          |        Decompressing
      Speed Usage    R/U Rating  |    Speed Usage    R/U Rating
       KB/s     %   MIPS   MIPS  |     KB/s     %   MIPS   MIPS

22:    2058   121   1651   2002  |    37927   178   1923   3424
23:    2161   129   1706   2202  |    35674   170   1921   3266
24:    2332   145   1732   2507  |    35250   171   1907   3271
25:    1994   131   1737   2276  |    34227   170   1888   3219
26:    1987   147   1643   2421  |    32412   245   1259   3088
22:    2150   126   1658   2091  |    36655   257   1286   3309
23:    1985   119   1695   2023  |    36117   172   1924   3307
24:    1954   122   1721   2101  |    34635   167   1927   3214
25:    2141   140   1745   2444  |    33862   167   1906   3185
26:    2023   151   1631   2465  |    32144   162   1885   3063
----------------------------------------------------------------
Avr:          133   1692   2253               186   1783   3234
Tot:          160   1737   2744


There are several other benchmarking tools. Check http://tldp.org (The Linux Documentation Project), http://google.com/linux and search with your package manager.

Feel free to post your benchmark results in the comments!

0 comments:

Post a Comment