You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@cassandra.apache.org by kevin <ke...@gmail.com> on 2010/02/23 18:33:06 UTC
reads are slow
to load 100 columns of a super column it takes over a second. how to improve
this performance.
i am using cassandra 0.5 version.
output of nodeprobe info
29814395632524962303611017038378268216
Load : 9.18 GB
Generation No : 1266945238
Uptime (seconds) : 638131
Heap Memory (MB) : 47.55 / 10237.94
the full config file is here
http://pastie.org/838916
I have given 10GB RAM in cassandra.in.sh.
-Xmx10G \
i have increased KeysCachedFraction to 0.04. i have two different drives for
commitlog and data directoy.
i have about 3 million rows.
what can i do to improve read speed?
thanks a lot!
Re: reads are slow
Posted by kevin <ke...@gmail.com>.
On Tue, Feb 23, 2010 at 11:49 AM, Jonathan Ellis <jb...@gmail.com> wrote:
> On Tue, Feb 23, 2010 at 12:12 PM, kevin <ke...@gmail.com>
> wrote:
> > On Tue, Feb 23, 2010 at 10:07 AM, Jonathan Ellis <jb...@gmail.com>
> wrote:
> >>
> >> you enable row caching by upgrading to 0.6. :)
> >
> > where can i get 0.6 from? svn trunk?
>
> svn branches/cassandra-0.6
>
> like I said, we're voting on a beta today, so you should probably just
> wait for that.
>
thanks!
i saw the changes and it says batch_insert is being deprecated, and a new
command batch_mutate is coming in. can you guys add example in the wiki API
page!
* add batch_mutate thrift command, deprecating batch_insert (CASSANDRA-336)
http://wiki.apache.org/cassandra/API
thanks a lot
Re: reads are slow
Posted by Jonathan Ellis <jb...@gmail.com>.
On Tue, Feb 23, 2010 at 12:12 PM, kevin <ke...@gmail.com> wrote:
> On Tue, Feb 23, 2010 at 10:07 AM, Jonathan Ellis <jb...@gmail.com> wrote:
>>
>> you enable row caching by upgrading to 0.6. :)
>
> where can i get 0.6 from? svn trunk?
svn branches/cassandra-0.6
like I said, we're voting on a beta today, so you should probably just
wait for that.
> are there any upgrade instructions from 0.5? do i just flush, stop and start
> cassandra?
pretty much. even the flush is optional. see "upgrade" section in
https://svn.apache.org/repos/asf/incubator/cassandra/branches/cassandra-0.6/NEWS.txt
Re: reads are slow
Posted by kevin <ke...@gmail.com>.
On Tue, Feb 23, 2010 at 10:07 AM, Jonathan Ellis <jb...@gmail.com> wrote:
> you enable row caching by upgrading to 0.6. :)
>
where can i get 0.6 from? svn trunk?
are there any upgrade instructions from 0.5? do i just flush, stop and start
cassandra?
thanks
Re: reads are slow
Posted by Jonathan Ellis <jb...@gmail.com>.
you enable row caching by upgrading to 0.6. :)
it would help, yes, but for really big rows like yours it's probably
not a silver bullet either.
On Tue, Feb 23, 2010 at 12:04 PM, kevin <ke...@gmail.com> wrote:
> On Tue, Feb 23, 2010 at 9:51 AM, Brandon Williams <dr...@gmail.com> wrote:
>>
>> On Tue, Feb 23, 2010 at 11:33 AM, kevin <ke...@gmail.com>
>> wrote:
>>>
>>> I have given 10GB RAM in cassandra.in.sh.
>>> -Xmx10G \
>>> i have increased KeysCachedFraction to 0.04. i have two different drives
>>> for commitlog and data directoy.
>>> i have about 3 million rows.
>>> what can i do to improve read speed?
>>> thanks a lot!
>>
>> Since 0.5 doesn't have row caching, 10G is probably too much to give the
>> JVM and is hampering the OS cache. Try between 2-6GB instead and see if
>> that helps.
>
> the machine has 24GB of RAM, and about 4G used and 17G cached.
> http://pastie.org/838979
>
> do you still think reducing the 10G will help? it seems to be a little
> counter-intuitive. i assumed giving JVM more RAM would help it!
>
> what is row caching? how do i enable it?
> thanks
>
Re: reads are slow
Posted by kevin <ke...@gmail.com>.
On Tue, Feb 23, 2010 at 9:51 AM, Brandon Williams <dr...@gmail.com> wrote:
> On Tue, Feb 23, 2010 at 11:33 AM, kevin <ke...@gmail.com>wrote:
>
>> I have given 10GB RAM in cassandra.in.sh.
>> -Xmx10G \
>> i have increased KeysCachedFraction to 0.04. i have two different drives
>> for commitlog and data directoy.
>> i have about 3 million rows.
>> what can i do to improve read speed?
>> thanks a lot!
>>
>
> Since 0.5 doesn't have row caching, 10G is probably too much to give the
> JVM and is hampering the OS cache. Try between 2-6GB instead and see if
> that helps.
>
the machine has 24GB of RAM, and about 4G used and 17G cached.
http://pastie.org/838979
do you still think reducing the 10G will help? it seems to be a little
counter-intuitive. i assumed giving JVM more RAM would help it!
what is row caching? how do i enable it?
thanks
Re: reads are slow
Posted by Brandon Williams <dr...@gmail.com>.
On Tue, Feb 23, 2010 at 11:33 AM, kevin <ke...@gmail.com> wrote:
> I have given 10GB RAM in cassandra.in.sh.
> -Xmx10G \
> i have increased KeysCachedFraction to 0.04. i have two different drives
> for commitlog and data directoy.
> i have about 3 million rows.
> what can i do to improve read speed?
> thanks a lot!
>
Since 0.5 doesn't have row caching, 10G is probably too much to give the JVM
and is hampering the OS cache. Try between 2-6GB instead and see if that
helps.
-Brandon
Re: reads are slow
Posted by Jonathan Ellis <jb...@gmail.com>.
no
On Tue, Feb 23, 2010 at 12:09 PM, kevin <ke...@gmail.com> wrote:
> On Tue, Feb 23, 2010 at 10:06 AM, Jonathan Ellis <jb...@gmail.com> wrote:
>>
>> the standard workaround is to change your data model to use non-super
>> columns instead.
>
> is there any limit on the number of standard column families that i can set
> up on cassandra?
>
Re: reads are slow
Posted by kevin <ke...@gmail.com>.
On Tue, Feb 23, 2010 at 10:06 AM, Jonathan Ellis <jb...@gmail.com> wrote:
> the standard workaround is to change your data model to use non-super
> columns instead.
>
is there any limit on the number of standard column families that i can set
up on cassandra?
Re: reads are slow
Posted by Jonathan Ellis <jb...@gmail.com>.
only the total row size limit (must fit in memory during compaction)
On Wed, Feb 24, 2010 at 7:47 AM, kevin <ke...@gmail.com> wrote:
> On Tue, Feb 23, 2010 at 10:06 AM, Jonathan Ellis <jb...@gmail.com> wrote:
>>
>> the standard workaround is to change your data model to use non-super
>> columns instead.
>>
>> supercolumns are really only for relatively small numbers of
>> subcolumns until 598 is addressed.
>
> is there any limit on the number of supercolumns i can have?
>
Re: reads are slow
Posted by kevin <ke...@gmail.com>.
On Tue, Feb 23, 2010 at 10:06 AM, Jonathan Ellis <jb...@gmail.com> wrote:
> the standard workaround is to change your data model to use non-super
> columns instead.
>
> supercolumns are really only for relatively small numbers of
> subcolumns until 598 is addressed.
>
is there any limit on the number of supercolumns i can have?
Re: reads are slow
Posted by Jonathan Ellis <jb...@gmail.com>.
the standard workaround is to change your data model to use non-super
columns instead.
supercolumns are really only for relatively small numbers of
subcolumns until 598 is addressed.
On Tue, Feb 23, 2010 at 12:01 PM, kevin <ke...@gmail.com> wrote:
> yes there is a lot more than 100. there are around 50,000 to 100000
> columns.
>
> is there any solution for this?
>
> from the issue 598 page, i can have only 1000s of columns in a super column?
>
> On Tue, Feb 23, 2010 at 9:47 AM, Jonathan Ellis <jb...@gmail.com> wrote:
>>
>> how many columns total are in the supercolumn?
>>
>> if it's a lot more than 100 you are running into
>> http://issues.apache.org/jira/browse/CASSANDRA-598
>>
>> On Tue, Feb 23, 2010 at 11:44 AM, kevin <ke...@gmail.com>
>> wrote:
>> > i dont think so
>> > /dev/sdc1 is the commitlog drive, and /dev/sdd1 is the data directory.
>> >
>> > http://pastie.org/838943
>> >
>> >
>> > On Tue, Feb 23, 2010 at 9:35 AM, Jonathan Ellis <jb...@gmail.com>
>> > wrote:
>> >>
>> >> are you i/o bound?
>> >> http://spyced.blogspot.com/2010/01/linux-performance-basics.html
>> >>
>> >> On Tue, Feb 23, 2010 at 11:33 AM, kevin <ke...@gmail.com>
>> >> wrote:
>> >> > to load 100 columns of a super column it takes over a second. how to
>> >> > improve
>> >> > this performance.
>> >> >
>> >> >
>> >> > i am using cassandra 0.5 version.
>> >> >
>> >> > output of nodeprobe info
>> >> > 29814395632524962303611017038378268216
>> >> > Load : 9.18 GB
>> >> > Generation No : 1266945238
>> >> > Uptime (seconds) : 638131
>> >> > Heap Memory (MB) : 47.55 / 10237.94
>> >> >
>> >> > the full config file is here
>> >> > http://pastie.org/838916
>> >> >
>> >> > I have given 10GB RAM in cassandra.in.sh.
>> >> > -Xmx10G \
>> >> > i have increased KeysCachedFraction to 0.04. i have two different
>> >> > drives
>> >> > for
>> >> > commitlog and data directoy.
>> >> > i have about 3 million rows.
>> >> > what can i do to improve read speed?
>> >> > thanks a lot!
>> >> >
>> >
>> >
>
>
Re: reads are slow
Posted by kevin <ke...@gmail.com>.
yes there is a lot more than 100. there are around 50,000 to 100000
columns.
is there any solution for this?
from the issue 598 page, i can have only 1000s of columns in a super column?
On Tue, Feb 23, 2010 at 9:47 AM, Jonathan Ellis <jb...@gmail.com> wrote:
> how many columns total are in the supercolumn?
>
> if it's a lot more than 100 you are running into
> http://issues.apache.org/jira/browse/CASSANDRA-598
>
> On Tue, Feb 23, 2010 at 11:44 AM, kevin <ke...@gmail.com>
> wrote:
> > i dont think so
> > /dev/sdc1 is the commitlog drive, and /dev/sdd1 is the data directory.
> >
> > http://pastie.org/838943
> >
> >
> > On Tue, Feb 23, 2010 at 9:35 AM, Jonathan Ellis <jb...@gmail.com>
> wrote:
> >>
> >> are you i/o bound?
> >> http://spyced.blogspot.com/2010/01/linux-performance-basics.html
> >>
> >> On Tue, Feb 23, 2010 at 11:33 AM, kevin <ke...@gmail.com>
> >> wrote:
> >> > to load 100 columns of a super column it takes over a second. how to
> >> > improve
> >> > this performance.
> >> >
> >> >
> >> > i am using cassandra 0.5 version.
> >> >
> >> > output of nodeprobe info
> >> > 29814395632524962303611017038378268216
> >> > Load : 9.18 GB
> >> > Generation No : 1266945238
> >> > Uptime (seconds) : 638131
> >> > Heap Memory (MB) : 47.55 / 10237.94
> >> >
> >> > the full config file is here
> >> > http://pastie.org/838916
> >> >
> >> > I have given 10GB RAM in cassandra.in.sh.
> >> > -Xmx10G \
> >> > i have increased KeysCachedFraction to 0.04. i have two different
> drives
> >> > for
> >> > commitlog and data directoy.
> >> > i have about 3 million rows.
> >> > what can i do to improve read speed?
> >> > thanks a lot!
> >> >
> >
> >
>
Re: reads are slow
Posted by Jonathan Ellis <jb...@gmail.com>.
how many columns total are in the supercolumn?
if it's a lot more than 100 you are running into
http://issues.apache.org/jira/browse/CASSANDRA-598
On Tue, Feb 23, 2010 at 11:44 AM, kevin <ke...@gmail.com> wrote:
> i dont think so
> /dev/sdc1 is the commitlog drive, and /dev/sdd1 is the data directory.
>
> http://pastie.org/838943
>
>
> On Tue, Feb 23, 2010 at 9:35 AM, Jonathan Ellis <jb...@gmail.com> wrote:
>>
>> are you i/o bound?
>> http://spyced.blogspot.com/2010/01/linux-performance-basics.html
>>
>> On Tue, Feb 23, 2010 at 11:33 AM, kevin <ke...@gmail.com>
>> wrote:
>> > to load 100 columns of a super column it takes over a second. how to
>> > improve
>> > this performance.
>> >
>> >
>> > i am using cassandra 0.5 version.
>> >
>> > output of nodeprobe info
>> > 29814395632524962303611017038378268216
>> > Load : 9.18 GB
>> > Generation No : 1266945238
>> > Uptime (seconds) : 638131
>> > Heap Memory (MB) : 47.55 / 10237.94
>> >
>> > the full config file is here
>> > http://pastie.org/838916
>> >
>> > I have given 10GB RAM in cassandra.in.sh.
>> > -Xmx10G \
>> > i have increased KeysCachedFraction to 0.04. i have two different drives
>> > for
>> > commitlog and data directoy.
>> > i have about 3 million rows.
>> > what can i do to improve read speed?
>> > thanks a lot!
>> >
>
>
Re: reads are slow
Posted by kevin <ke...@gmail.com>.
i dont think so
/dev/sdc1 is the commitlog drive, and /dev/sdd1 is the data directory.
http://pastie.org/838943
On Tue, Feb 23, 2010 at 9:35 AM, Jonathan Ellis <jb...@gmail.com> wrote:
> are you i/o bound?
> http://spyced.blogspot.com/2010/01/linux-performance-basics.html
>
> On Tue, Feb 23, 2010 at 11:33 AM, kevin <ke...@gmail.com>
> wrote:
> > to load 100 columns of a super column it takes over a second. how to
> improve
> > this performance.
> >
> >
> > i am using cassandra 0.5 version.
> >
> > output of nodeprobe info
> > 29814395632524962303611017038378268216
> > Load : 9.18 GB
> > Generation No : 1266945238
> > Uptime (seconds) : 638131
> > Heap Memory (MB) : 47.55 / 10237.94
> >
> > the full config file is here
> > http://pastie.org/838916
> >
> > I have given 10GB RAM in cassandra.in.sh.
> > -Xmx10G \
> > i have increased KeysCachedFraction to 0.04. i have two different drives
> for
> > commitlog and data directoy.
> > i have about 3 million rows.
> > what can i do to improve read speed?
> > thanks a lot!
> >
>
Re: reads are slow
Posted by Jonathan Ellis <jb...@gmail.com>.
are you i/o bound?
http://spyced.blogspot.com/2010/01/linux-performance-basics.html
On Tue, Feb 23, 2010 at 11:33 AM, kevin <ke...@gmail.com> wrote:
> to load 100 columns of a super column it takes over a second. how to improve
> this performance.
>
>
> i am using cassandra 0.5 version.
>
> output of nodeprobe info
> 29814395632524962303611017038378268216
> Load : 9.18 GB
> Generation No : 1266945238
> Uptime (seconds) : 638131
> Heap Memory (MB) : 47.55 / 10237.94
>
> the full config file is here
> http://pastie.org/838916
>
> I have given 10GB RAM in cassandra.in.sh.
> -Xmx10G \
> i have increased KeysCachedFraction to 0.04. i have two different drives for
> commitlog and data directoy.
> i have about 3 million rows.
> what can i do to improve read speed?
> thanks a lot!
>