New faceting implementation for 1.0 - planning resources when using 0.90.x now

Hi,

I am interested in learning more about planned new faceting
implementation(s) that might appear in ES 1.0 release.
All I know about new faceting implementation is based on the following
comments:


More specifically, if I plan to go to production with 0.90.x is there any
recommendation or estimate about how much resources I need to keep around
to allow for easy upgrade to 1.0 later?

Can we expect that there will be added a new facet implementation(s) next
to existing one or the new will replace the old? What kind of impact this
might have on memory and index size requirements when using some of
existing facets after 1.0 is released?

Regards,
Lukas

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

I referenced the same comment after I read it as well:
https://groups.google.com/forum/?fromgroups=#!topic/elasticsearch/HfbdBvi91dA

I'm guessing that the Lucene 4 transition opened up a few new avenues for
facets. I have been holding off writing my own facets since I do not want
to do work twice.

--
Ivan

On Thu, May 9, 2013 at 5:10 AM, Lukáš Vlček lukas.vlcek@gmail.com wrote:

Hi,

I am interested in learning more about planned new faceting
implementation(s) that might appear in ES 1.0 release.
All I know about new faceting implementation is based on the following
comments:

https://github.com/elasticsearch/elasticsearch/issues/1076#issuecomment-17566680

https://github.com/elasticsearch/elasticsearch/issues/1076#issuecomment-17636880

More specifically, if I plan to go to production with 0.90.x is there any
recommendation or estimate about how much resources I need to keep around
to allow for easy upgrade to 1.0 later?

Can we expect that there will be added a new facet implementation(s) next
to existing one or the new will replace the old? What kind of impact this
might have on memory and index size requirements when using some of
existing facets after 1.0 is released?

Regards,
Lukas

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

Ivan,

implementing custom facets is one thing but I am more into planning HW
resources for the future. For example, I want to go into production with
0.90.x and I need to specify my HW requirements now. I can do my best to
plan it based on current 0.90 performance. But I would like to account for
planned upgrade to 1.0 in the future without need to request servers with
different setup.

All I would like to know is if there is any recommendation or estimate in
this regard.

Regards,
Lukas

On Thu, May 9, 2013 at 5:34 PM, Ivan Brusic ivan@brusic.com wrote:

I referenced the same comment after I read it as well:
https://groups.google.com/forum/?fromgroups=#!topic/elasticsearch/HfbdBvi91dA

I'm guessing that the Lucene 4 transition opened up a few new avenues for
facets. I have been holding off writing my own facets since I do not want
to do work twice.

--
Ivan

On Thu, May 9, 2013 at 5:10 AM, Lukáš Vlček lukas.vlcek@gmail.com wrote:

Hi,

I am interested in learning more about planned new faceting
implementation(s) that might appear in ES 1.0 release.
All I know about new faceting implementation is based on the following
comments:

https://github.com/elasticsearch/elasticsearch/issues/1076#issuecomment-17566680

https://github.com/elasticsearch/elasticsearch/issues/1076#issuecomment-17636880

More specifically, if I plan to go to production with 0.90.x is there any
recommendation or estimate about how much resources I need to keep around
to allow for easy upgrade to 1.0 later?

Can we expect that there will be added a new facet implementation(s) next
to existing one or the new will replace the old? What kind of impact this
might have on memory and index size requirements when using some of
existing facets after 1.0 is released?

Regards,
Lukas

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

I cannot imagine that hardware requirements would be increased by any new
facet implementations. If anything, leveraging the new Lucene features such
as DocValues will only improve things. v0.90.x should already be a huge
improvement (if you fall under the high-cardinality use case) although I
have yet to do any testing.

My comment about my comment was to simply state that I also have questions
about the 1.0 release and it would be great if the team can chime in.

--
Ivan

On Thu, May 9, 2013 at 10:38 AM, Lukáš Vlček lukas.vlcek@gmail.com wrote:

Ivan,

implementing custom facets is one thing but I am more into planning HW
resources for the future. For example, I want to go into production with
0.90.x and I need to specify my HW requirements now. I can do my best to
plan it based on current 0.90 performance. But I would like to account for
planned upgrade to 1.0 in the future without need to request servers with
different setup.

All I would like to know is if there is any recommendation or estimate in
this regard.

Regards,
Lukas

On Thu, May 9, 2013 at 5:34 PM, Ivan Brusic ivan@brusic.com wrote:

I referenced the same comment after I read it as well:
https://groups.google.com/forum/?fromgroups=#!topic/elasticsearch/HfbdBvi91dA

I'm guessing that the Lucene 4 transition opened up a few new avenues for
facets. I have been holding off writing my own facets since I do not want
to do work twice.

--
Ivan

On Thu, May 9, 2013 at 5:10 AM, Lukáš Vlček lukas.vlcek@gmail.comwrote:

Hi,

I am interested in learning more about planned new faceting
implementation(s) that might appear in ES 1.0 release.
All I know about new faceting implementation is based on the following
comments:

https://github.com/elasticsearch/elasticsearch/issues/1076#issuecomment-17566680

https://github.com/elasticsearch/elasticsearch/issues/1076#issuecomment-17636880

More specifically, if I plan to go to production with 0.90.x is there
any recommendation or estimate about how much resources I need to keep
around to allow for easy upgrade to 1.0 later?

Can we expect that there will be added a new facet implementation(s)
next to existing one or the new will replace the old? What kind of impact
this might have on memory and index size requirements when using some of
existing facets after 1.0 is released?

Regards,
Lukas

--
You received this message because you are subscribed to the Google
Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send
an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

+1

I would like to know how/when to start implementing
equidistant/non-equidistant date histograms. Not that I can promise an
implementation now, but I'm in a state of limbo because I am also
obligated to migrate to ES 1.0 as soon as it's there.

Jörg

Am 09.05.13 17:34, schrieb Ivan Brusic:

I'm guessing that the Lucene 4 transition opened up a few new avenues
for facets. I have been holding off writing my own facets since I do
not want to do work twice.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

I am sorry if it is not appropriate to dig an old message again, but I also
want to find out what will happen to the 1.0 and can just come up with this
message and also want to hear if there is any update about the questions
mentioned here.

I am also in a stage that want to launch a new elasticsearch cluster to
production. We have started the development for nearly 1 year starting from
version 0.19.* and just upgrade to 0.90.*

But if 1.0 is near to release, I also want to migrate to 1.0 directly
before putting my new elasticsearch cluster to production.

On Friday, May 10, 2013 3:49:36 AM UTC+8, Jörg Prante wrote:

+1

I would like to know how/when to start implementing
equidistant/non-equidistant date histograms. Not that I can promise an
implementation now, but I'm in a state of limbo because I am also
obligated to migrate to ES 1.0 as soon as it's there.

Jörg

Am 09.05.13 17:34, schrieb Ivan Brusic:

I'm guessing that the Lucene 4 transition opened up a few new avenues
for facets. I have been holding off writing my own facets since I do
not want to do work twice.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

Hi,

I believe that https://github.com/elasticsearch/elasticsearch/issues/3300is
related to this question. (kudos to @uboness for this detailed
writeup)

The way I understand it is that ES 1.0.0 will likely contain existing
[legacy?] faceting and new aggregations code as well. So I would say that
if your (or mine) code sticks to existing faceting then we should
[hopefully] not see any significant memory/perf changes once 1.0.0 is
released. Only if we switch to new aggregations then the things could
change. I would expect that it will take some time for the aggregations
code to settle. But definitely aggregations look very cool.

Regards,
Lukas

On Thu, Jul 18, 2013 at 12:10 PM, Yiu Wing TSANG ywtsang@gmail.com wrote:

I am sorry if it is not appropriate to dig an old message again, but I
also want to find out what will happen to the 1.0 and can just come up with
this message and also want to hear if there is any update about the
questions mentioned here.

I am also in a stage that want to launch a new elasticsearch cluster to
production. We have started the development for nearly 1 year starting from
version 0.19.* and just upgrade to 0.90.*

But if 1.0 is near to release, I also want to migrate to 1.0 directly
before putting my new elasticsearch cluster to production.

On Friday, May 10, 2013 3:49:36 AM UTC+8, Jörg Prante wrote:

+1

I would like to know how/when to start implementing
equidistant/non-equidistant date histograms. Not that I can promise an
implementation now, but I'm in a state of limbo because I am also
obligated to migrate to ES 1.0 as soon as it's there.

Jörg

Am 09.05.13 17:34, schrieb Ivan Brusic:

I'm guessing that the Lucene 4 transition opened up a few new avenues
for facets. I have been holding off writing my own facets since I do
not want to do work twice.

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.