search speed

[gongtao]

there are 300,000 records in the repository.

Why the increase as the result set, for speed will slow ? I set up to offset 15   limit 100

Statements:
//nko:workspace/*[@jcr:uuid='a2e5266f-9f2f-4683-993f-b73f34ba94a7' or @jcr:uuid='5fbc094f-e6c6-48b3-9083-49fe18a27c4a']//element(*,nt:unstructured)[@templateName='4137345d-3acd-4fc6-befe-6367a3c8f842' and @stateCode=1  and  @directory=false()] order by @name ascending

results:30,000  time:5703ms
results:300,000  time:30406ms



Full Text Search more slowly
Statements:
/jcr:root/nko:root/nko:workspace/*[@jcr:uuid='a2e5266f-9f2f-4683-993f-b73f34ba94a7' or @jcr:uuid='5fbc094f-e6c6-48b3-9083-49fe18a27c4a']//element(*,nt:base)[jcr:contains(.,'superman OR *superman*') and  @stateCode=1 and @directory=false()] order by @lastModified ascending

results:30,000 time:14360ms

[Ard Schrijvers-3]

Hello gongtao,

I do not get this sentence 'Why the increase as the result set, for
speed will slow ? ' though you might want to take a look at [1]

A little strange perhapes, but if you replace

'//nko:workspace/*' with //* (you are testing for a uuid anyway) you
might gain a lot of speed. Hierarchy information in a query is costly,
certainly for the first hit. So also, your performance numbers do nto
tell me very much, because there is a huge difference between first run
queries and consecutive queries where caches, and then mainly the
hierchical caches, are warmed up.

Anyway, if you read [1] you should be able to understand a lot more
about the queries you are doing,

-Ard

[1]
http://www.nabble.com/Explanation-and-solutions-of-some-Jackrabbit-queri
es-regarding-performance-td15028655.html

>
>
> there are 300,000 records in the repository.
>
> Why the increase as the result set, for speed will slow ? I
> set up to offset
> 15   limit 100
>
> Statements:
> //nko:workspace/*[@jcr:uuid='a2e5266f-9f2f-4683-993f-b73f34ba9

4a7' or @jcr:uuid='5fbc094f->
e6c6-48b3-9083-49fe18a27c4a']//element(*,nt:unstructured)[@tem

> plateName='4137345d-3acd-4fc6-befe-6367a3c8f842'
> and @stateCode=1  and  @directory=false()] order by @name ascending
>
> results:30,000  time:5703ms
> results:300,000  time:30406ms
>
>
>
> Full Text Search more slowly
> Statements:
> /jcr:root/nko:root/nko:workspace/*[@jcr:uuid='a2e5266f-9f2f-46

83-993f-b73f34ba94a7'
> or
> @jcr:uuid='5fbc094f-e6c6-48b3-9083-49fe18a27c4a']//element(*,n
t:base)[jcr:contains(.,'superman

> OR *superman*') and  @stateCode=1 and @directory=false()]
> order by @lastModified ascending
>
> results:30,000 time:14360ms
> --
> View this message in context:
> http://www.nabble.com/search-speed-tp18583171p18583171.html
> Sent from the Jackrabbit - Users mailing list archive at Nabble.com.
>
>

[gongtao]

Hi Ard Schrijvers-3,

I tried your way ,query speed  the four seconds faster than before.
However, full-text search, retrieve version information
the full-text search  Statements:
 //*[@jcr:uuid='a2e5266f-9f2f-4683-993f-b73f34ba94a7' or @jcr:uuid='5fbc094f-e6c6-48b3-908349fe18a27c4a']//element(*,nt:base)[jcr:contains(.,'superman OR *superman*') and  @stateCode=1 and @directory=false()] order by @lastModified ascending

Ard Schrijvers-3 wrote:

Hello gongtao,

I do not get this sentence 'Why the increase as the result set, for
speed will slow ? ' though you might want to take a look at [1]

A little strange perhapes, but if you replace

'//nko:workspace/*' with //* (you are testing for a uuid anyway) you
might gain a lot of speed. Hierarchy information in a query is costly,
certainly for the first hit. So also, your performance numbers do nto
tell me very much, because there is a huge difference between first run
queries and consecutive queries where caches, and then mainly the
hierchical caches, are warmed up.

Anyway, if you read [1] you should be able to understand a lot more
about the queries you are doing,

-Ard

[1]
http://www.nabble.com/Explanation-and-solutions-of-some-Jackrabbit-queri
es-regarding-performance-td15028655.html

>
>
> there are 300,000 records in the repository.
>
> Why the increase as the result set, for speed will slow ? I
> set up to offset
> 15   limit 100
>
> Statements:
> //nko:workspace/*[@jcr:uuid='a2e5266f-9f2f-4683-993f-b73f34ba9
4a7' or @jcr:uuid='5fbc094f->
e6c6-48b3-9083-49fe18a27c4a']//element(*,nt:unstructured)[@tem
> plateName='4137345d-3acd-4fc6-befe-6367a3c8f842'
> and @stateCode=1  and  @directory=false()] order by @name ascending
>
> results:30,000  time:5703ms
> results:300,000  time:30406ms
>
>
>
> Full Text Search more slowly
> Statements:
> /jcr:root/nko:root/nko:workspace/*[@jcr:uuid='a2e5266f-9f2f-46
83-993f-b73f34ba94a7'
> or
> @jcr:uuid='5fbc094f-e6c6-48b3-9083-49fe18a27c4a']//element(*,n
t:base)[jcr:contains(.,'superman
> OR *superman*') and  @stateCode=1 and @directory=false()]
> order by @lastModified ascending
>
> results:30,000 time:14360ms
> --
> View this message in context:
> http://www.nabble.com/search-speed-tp18583171p18583171.html
> Sent from the Jackrabbit - Users mailing list archive at Nabble.com.
>
>