Performance Hit With Many Files Post: 302294978

Sponsored Content

Special Forums Hardware Filesystems, Disks and Memory Performance Hit With Many Files Post 302294978 by MarkSeger on Friday 6th of March 2009 08:59:24 AM

03-06-2009

Registered User

I too have done similar testing but slightly different. Rather than simply time the creation of 100 files, which can be very misleading OR timing the creation of 1M files which is no better, I prefer to look at what is happening across the system resources during the entire event.

I run collectl with a monitoring interval of 1 second, logging to a file or simply watching the system in real time. When I create a million files I can watch the cpu periodically increase. In fact, when getting in the higher ends of files I can actually see spike in cpu load. This is something you can't see when just doing end-to-end numbers.

Another interesting test is to set up an alarm in your script to write out the number of files created every 10th (or even hundredth) of a second. You'll be amazed to see how linearly the number of files created/second drops over time as well as how things periodically slow down but are not visible when only looking at second-level samples.

You can also run collectl at a monitoring interval of 0.1 seconds and see micro-spikes in CPU load as well. This is something most people miss because none of the existing tools can deal with sub-second reporting.

-mark

MarkSeger

View Public Profile for MarkSeger

Find all posts by MarkSeger

9 More Discussions You Might Find Interesting

1. UNIX for Dummies Questions & Answers

Anyone else see a performance hit from ext3

I reinstalled my Linux box with RedHat 7.2 and used the ext3 journaling file system. This thing is a pig now. There isn't much running on the box, and performance is sad.

2. Shell Programming and Scripting

Shell Script to hit a url

Hi all, I have a php file that grabs xml, parses it and updates my db accordingly. I want to automate the execution of this process, rather than having to hit the url manually. I have been looking into using cron to execute a script to do this, however i'm not exactly sure what command i would...

3. Programming

why multiple SIGINT raises when i hit C-c

hi, in my application, i have set up to capture SIGINT and execute a handler.the problem is whenever i hit C-c, multiple SIGINT are sent to the application.I have blocked the SIGINT right after catching the first one but it is unsuccessful.Here is what i do : jmp_buf main_loop; int...

4. Shell Programming and Scripting

Hit count on a shell script

I have a unix shell script (ex.sh) written. How to find out how many users (incl. myself) have run this .sh ? I can insert code snipet at top of script if need be. - Ravi

5. Shell Programming and Scripting

Getting Next Best Hit..

Hi.. I need to get the following output from the input file like this INPUT GRM1 GRM1 0 GRM1 ABC1 1 GRM1 FEQ1 2 GRM1 SED1 3 ABC2 GRM1 0 ABC2 ABC2 1 ABC2 FEQ1 2 ABC2 BED1 3 SED1 SED1 0 SED1 SED1 1 SED1 SED1 2 SED1 ABC1 3 OUTPUT:

6. SuSE

Java hit

Hello, I'm having trouble looking for info for SUSIE on this CVE-2012-4681. This is basically the newest Java hit. It is mostly a web browser issue but I would like to see if the versions on our servers are vulnerable. I already found the pages/info for Solaris and RHEL. Any help would be...

7. Cybersecurity

vnc password hit from Retina

Hello, I'm having an issue with VNC. Security at work says that they scanned my servers (Solaris, RHEL, SLES) and found that you don't need a password to access a VNC session. I have tested this and you can't login to the VNC session without a password. Can someone tell what the Retina scanner...

8. Shell Programming and Scripting

Curl to hit the submit button

Hello, I am looking to hit a URL using curl and click on submit button so that I can get the results. The below is the code <input name="tos_accepted" id="tos_accepted" class="button" type="submit" value="Yes, I Agree"/> <input name="tos_discarded" id="tos_discarded"...

9. Shell Programming and Scripting

Best performance to merge two files

Hi Gurus, I need to merge two files. file1 (small file, only one line) this is first linefile2 (large file) abc def ghi ... I use below command to merge the file, since the file2 is really large file, the command read whole file2, the performance is not good. cat file1 > file3...

LEARN ABOUT DEBIAN

scrounge-ntfs

scrounge-ntfs(8)					    BSD System Manager's Manual 					  scrounge-ntfs(8)

NAME

     scrounge-ntfs -- helps retrieve data from corrupted NTFS partitions

SYNOPSIS

     scrounge-ntfs -l disk
     scrounge-ntfs -s disk
     scrounge-ntfs [-m mftoffset] [-c clustersize] [-o outdir] disk start end

DESCRIPTION

     scrounge-ntfs is a utility that can rescue data from corrupted NTFS partitions. It writes the files retrieved to another working file system.
     Certain information about the partition needs to be known in advance.

     The -l mode is meant to be run in advance of the data corruption, with the output stored away in a file. This allows scrounge-ntfs to recover
     data reliably. See the 'NOTES' section below for recover info when this isn't the case.

OPTIONS

     The options are as follows:

     -c 	 The cluster size (in sectors). When not specified a default of 8 is used.

     -l 	 List partition information for a drive. This will only work when the partition table for the given drive is intact.

     -m 	 When recovering data this specifies the location of the MFT from the beginning of the partition (in sectors). If not specified
		 then no directory information can be used, that is, all rescued files will be written to the same directory.

     -o 	 Directory to put rescued files in. If not specified then files will be placed in the current directory.

     -s 	 Search disk for partition information. (Not implemented yet).

     disk	 The raw device used to access the disk which contains the NTFS partition to rescue files from. eg: '/dev/hdc'

     start	 The beginning of the NTFS partition (in sectors).

     end	 The end of the NTFS partition (in sectors)

NOTES

     If you plan on using this program sucessfully you should prepare in advance by storing a copy of the partition information. Use the -l option
     to do this. Eventually searching for disk partition information will be implemented, which will solve this problem.

     When only one partition exists on a disk or you want to rescue the first partition there are ways to guess at the sector sizes and MFT loca-
     tion. See the scrounge-ntfs web page for more info:

     http://memberwebs.com/swalter/software/scrounge/

AUTHOR

     Stef Walter <stef@memberwebs.com>

scrounge-ntfs							   June 1, 2019 						     scrounge-ntfs