11-12-2008
AFAIK there is no such thing with DIO/CIO to improve throughput of large sequential reads like jfs/jfs2 do with their read-ahead option. I.e. DIO/CIO even may deteriorate performance of applications that read data most of the time.
8 More Discussions You Might Find Interesting
1. HP-UX
One cool thing about unix is that it predicts disk blocks that you may need and tries to have them in core before you need them. Over the years, various unix vendors tried various algorithms to improve performance. HP has patented their latest algorithm...
Multi-threaded Read Ahead Prediction... (0 Replies)
Discussion started by: Perderabo
0 Replies
2. AIX
hi all,
can sumbody give me a link which gives the basic layout of JFS, JFS2 and the veritas file system.
and i also want to know about the data structures used in this filsystem
thanx in advance (0 Replies)
Discussion started by: anwerreyaz
0 Replies
3. Shell Programming and Scripting
I would like to search a router config file for "ip address $ip", once found, I want to grab the line just before that contains "interface $interfacetype"
basically saying, 10.3.127.9 is assigned to "Loopback1" given the below as an example.
interface Loopback1
ip address 10.3.127.9... (1 Reply)
Discussion started by: popeye
1 Replies
4. AIX
Hi all,
Can anyone define the difference between jfs and jfs2 filesystem as well as usage of jfs log files....
Thanks.... (7 Replies)
Discussion started by: sumathi.k
7 Replies
5. Red Hat
Hi Friends,
Please help me to understand, how to enable async disk IO and Direct disk IO in ext3 filesystem on rhel5.
Regards,
Arumon (0 Replies)
Discussion started by: arumon
0 Replies
6. AIX
Hi,
I have a bit of a NFS problem on AiX 6.1 : When I set the mount to cio and dio - needed for a database app - Everything slows down. The following is copying 700mb, top one is a normal mount bottom one is a mount with the cio/dio option enabled :
# ./a.sh
Wed Jan 11 11:41:24 GMT 2012... (5 Replies)
Discussion started by: AJCG1976
5 Replies
7. AIX
Hello!
I have a LPAR with 51GB RAM
topas shows 21% of memory for client pages (~11Gb)
svmon -G show 2906772 client pages in use (~11Gb)
but then i try to investigate per process client memory usage
svmon -P -O filtertype=client
and summarize inuse column, i get only 347880 pages... (1 Reply)
Discussion started by: sys
1 Replies
8. UNIX for Beginners Questions & Answers
Hi,
I have a code that will show one day ahead, how to make it show two days ahead
p-dev1-db-tst:/$ day=$(TZ=IST-24 date +%d)
p-dev1-db-tst:/$ echo $day
17
p-dev1-db-tst:/$
Regards, Adam (5 Replies)
Discussion started by: answer
5 Replies
LEARN ABOUT OPENSOLARIS
sd_readahead
SD_READAHEAD(3) sd_readahead SD_READAHEAD(3)
NAME
sd_readahead - Control ongoing disk boot-time read-ahead operations
SYNOPSIS
#include "sd-readahead.h"
int sd_readahead(const char *action);
DESCRIPTION
sd_readahead() may be called by programs involved with early boot-up to control ongoing boot-time disk read-ahead operations. It may be
used to terminate read-ahead operations in case an uncommon disk access pattern is to be expected and hence read-ahead replay or collection
is unlikely to have the desired speed-up effect on the current or future boot-ups.
The action should be one of the following strings:
cancel
Terminates read-ahead data collection, and drops all read-ahead data collected during this boot-up.
done
Terminates read-ahead data collection, but keeps all read-ahead data collected during this boot-up around for use during subsequent
boot-ups.
noreplay
Terminates read-ahead replay.
RETURN VALUE
On failure, these calls return a negative errno-style error code. It is generally recommended to ignore the return value of this call.
NOTES
This function is provided by the reference implementation of APIs for controlling boot-time read-ahead and distributed with the systemd
package. The algorithm it implements is simple, and can easily be reimplemented in daemons if it is important to support this interface
without using the reference implementation.
Internally, this function creates a file in /run/systemd/readahead/ which is then used as flag file to notify the read-ahead subsystem.
For details about the algorithm check the liberally licensed reference implementation sources:
http://cgit.freedesktop.org/systemd/systemd/plain/src/readahead/sd-readahead.c resp.
http://cgit.freedesktop.org/systemd/systemd/plain/src/systemd/sd-readahead.h
sd_readahead() is implemented in the reference implementation's drop-in sd-readahead.c and sd-readahead.h files. It is recommended that
applications consuming this API copy the implementation into their source tree. For more details about the reference implementation see sd-
readahead(7)
If -DDISABLE_SYSTEMD is set during compilation this function will always return 0 and otherwise become a NOP.
EXAMPLES
Example 1. Cancelling all read-ahead operations
During boots where SELinux has to relabel the file system hierarchy, it will create a large amount of disk accesses that are not necessary
during normal boots. Hence it is a good idea to disable both read-ahead replay and read-ahead collection.
sd_readahead("cancel");
sd_readahead("noreplay");
SEE ALSO
systemd(1), sd-readahead(7), daemon(7)
AUTHOR
Lennart Poettering <lennart@poettering.net>
Developer
systemd 10/07/2013 SD_READAHEAD(3)