mongo.pl PHASE_* options question

mongo.pl PHASE_* options question

[David Dabbs]

I’d like to reproduce the mongo tests as shown at 

	http://www.namesys.com/benchmarks/v4marks.html

The page presents the mongo options used to generate the results, but not
all of these are documented at 

	http://www.namesys.com/benchmarks/mongo_readme.html

,and they are not mentioned in the mongo distribution README.


FSTYPE= ext3 | reiserfs | reiser4
MKFS= mkfs.<FSTYPE>

DEV=/dev/hda2   (or whatever disk device) 
DIR=/mnt/testfs (mountpoint you must create)
FILE_SIZE=4000 
BYTES=512000000 
#
# To have results free from buffer cache influence, 
# settings must satisfy 
#
#    BYTES * REP_COUNTER > ramsize
#

MOUNT_OPTIONS=<fs-specific mount options passed to MKFS command>

NPROC=1
REP_COUNTER=1 
SYNC=off 
WRITE_BUFFER=131072 
GAMMA=0.2 

DD_MBCOUNT=768

#
# these are not documented
#
PHASE_COPY=cp 
PHASE_OVERWRITE=off 
PHASE_APPEND=off 
PHASE_READ=find 
PHASE_DELETE=rm 
PHASE_MODIFY=off


A quick scan of mongo.pl code that validates options shows that the PHASE_*
options are not present, so it is safe to assume that the PHASE* options are
defunct. If this is the case, perhaps these options should be removed from
the page.

David


# description of all valid options which can be in an input file is a
# hash array each element of it has the following form: 
# "key" => ["default value", flags, comment string]

%OPTIONS = (
"FSTYPE"  => [ undef, $SHOULD_BE_SET, "filesystem type"],
"DIR"     => [ undef, $SHOULD_BE_SET, "fs mount point"],
"MKFS"    => [ undef, 0, "mkfs command"],
"DEV"     => [ undef, $SHOULD_BE_SET, "device file name"],
"JOURNAL_DEV"  => [ undef, 0, "journal device file name"],
"JOURNAL_SIZE" => [ undef, 0, "journal size in 4K blocks"],
"MOUNT_OPTIONS" => [ undef, 0, "mount options"],
"BYTES" => [ undef, $SHOULD_BE_SET, "file set size (in bytes) created by all
instances of reiser_fract_tree in one pass, "],
"FILE_SIZE" => [ undef, $SHOULD_BE_SET, "median file size (a
reiser_fract_tree parameter)"],
"REP_COUNTER" => [3, 0, "number of passes for each mongo phase"],
"NPROC" => [3, 0, "munber of processes running simultaneously"],
"WRITE_BUFFER" => [4096, 0, "read/write buffer size parameter for mongo
utilities"], 
"GAMMA" => ["0.0", 0, "the exponent of core distribution in
reiser_fract_tree"],
"SYNC" => [off, 0, "using O_SYNC/fsync options to iozone"],
"DD_MBCOUNT" => [undef, 0, "size of largefile in megabytes"]
 
 

RE: mongo.pl PHASE_* options question

[David Dabbs]

> -----Original Message-----
> From: Edward Shishkin [mailto:edward@namesys.com]
> Sent: Friday, July 23, 2004 4:09 AM
> To: David Dabbs; reiserfs-dev@namesys.com
> Cc: grev@namesys.com
> Subject: Re: mongo.pl PHASE_* options question
> 

===SNIP ===

> 
> olalala, Lena would you please add those new options to
> http://www.namesys.com/benchmarks/mongo_readme.html
> and update the package this page references to (I'll be absent
> next week). Thanks, David.
> 

[David Dabbs] 
Glad I could help! BTW, could someone please send notice to the
reiserfs-list when the mongo package is updated? Thanks much,

dd

Re: mongo.pl PHASE_* options question

[E. Gryaznova]

Hello,

unfortunately this is the very old mongo version. Please try the new one :

http://www.namesys.com/benchmarks/mongo-2003.06.24.tar.gz

which has an ability to selectively turn phases on/off.

The new options documentation will be added tomorrow.

Thanks,
Lena.

David Dabbs wrote:

>I’d like to reproduce the mongo tests as shown at 
>
>	http://www.namesys.com/benchmarks/v4marks.html
>
>The page presents the mongo options used to generate the results, but not
>all of these are documented at 
>
>	http://www.namesys.com/benchmarks/mongo_readme.html
>
>,and they are not mentioned in the mongo distribution README.
>
>
>FSTYPE= ext3 | reiserfs | reiser4
>MKFS= mkfs.<FSTYPE>
>
>DEV=/dev/hda2   (or whatever disk device) 
>DIR=/mnt/testfs (mountpoint you must create)
>FILE_SIZE=4000 
>BYTES=512000000 
>#
># To have results free from buffer cache influence, 
># settings must satisfy 
>#
>#    BYTES * REP_COUNTER > ramsize
>#
>
>MOUNT_OPTIONS=<fs-specific mount options passed to MKFS command>
>
>NPROC=1
>REP_COUNTER=1 
>SYNC=off 
>WRITE_BUFFER=131072 
>GAMMA=0.2 
>
>DD_MBCOUNT=768
>
>#
># these are not documented
>#
>PHASE_COPY=cp 
>PHASE_OVERWRITE=off 
>PHASE_APPEND=off 
>PHASE_READ=find 
>PHASE_DELETE=rm 
>PHASE_MODIFY=off
>
>
>A quick scan of mongo.pl code that validates options shows that the PHASE_*
>options are not present, so it is safe to assume that the PHASE* options are
>defunct. If this is the case, perhaps these options should be removed from
>the page.
>
>David
>
>
># description of all valid options which can be in an input file is a
># hash array each element of it has the following form: 
># "key" => ["default value", flags, comment string]
>
>%OPTIONS = (
>"FSTYPE"  => [ undef, $SHOULD_BE_SET, "filesystem type"],
>"DIR"     => [ undef, $SHOULD_BE_SET, "fs mount point"],
>"MKFS"    => [ undef, 0, "mkfs command"],
>"DEV"     => [ undef, $SHOULD_BE_SET, "device file name"],
>"JOURNAL_DEV"  => [ undef, 0, "journal device file name"],
>"JOURNAL_SIZE" => [ undef, 0, "journal size in 4K blocks"],
>"MOUNT_OPTIONS" => [ undef, 0, "mount options"],
>"BYTES" => [ undef, $SHOULD_BE_SET, "file set size (in bytes) created by all
>instances of reiser_fract_tree in one pass, "],
>"FILE_SIZE" => [ undef, $SHOULD_BE_SET, "median file size (a
>reiser_fract_tree parameter)"],
>"REP_COUNTER" => [3, 0, "number of passes for each mongo phase"],
>"NPROC" => [3, 0, "munber of processes running simultaneously"],
>"WRITE_BUFFER" => [4096, 0, "read/write buffer size parameter for mongo
>utilities"], 
>"GAMMA" => ["0.0", 0, "the exponent of core distribution in
>reiser_fract_tree"],
>"SYNC" => [off, 0, "using O_SYNC/fsync options to iozone"],
>"DD_MBCOUNT" => [undef, 0, "size of largefile in megabytes"]
> 
> 
>
>
>
>
>
>  
>