You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@jena.apache.org by Steve Vestal <st...@adventiumlabs.com> on 2021/04/10 15:47:34 UTC
ARQ start-up overheads?
When doing some bench-marking, I ran the same query twice back-to-back.
The first time took about 4000ms, the second about 700ms. The
difference remained with optimizations on or off. This seems to happen
only on the first query, successive pairs of identical queries don't
show this effect. Is there some sort of start-up overhead for the very
first query issued?
Re: ARQ start-up overheads?
Posted by Andy Seaborne <an...@apache.org>.
On 10/04/2021 17:05, Steve Vestal wrote:
> It occurred to me, I am issuing the queries to an OntModel that is using
> openllet. Perhaps the reasoner is doing some start-up work.
>
> On 4/10/2021 10:47 AM, Steve Vestal wrote:
>> When doing some bench-marking, I ran the same query twice
>> back-to-back. The first time took about 4000ms, the second about
>> 700ms. The difference remained with optimizations on or off. This
>
>> seems to happen only on the first query, successive pairs of identical
>> queries don't show this effect. Is there some sort of start-up
>> overhead for the very first query issued?
>>
>>
>
Start-up:
Java - both reading in byte code and JIT if running the query in the
same JVM. If you run enough queries in the same JVM, you will see JIT
effects but it takes a few queries to start to kick in.
Data - if it is on-disk, the first time will bring it into the file
system cache. This also applies to bytecode in jar files.
Database - as data. The OS file system cache is significant.
And the reasoner.
Andy
Re: ARQ start-up overheads?
Posted by Steve Vestal <st...@adventiumlabs.com>.
It occurred to me, I am issuing the queries to an OntModel that is using
openllet. Perhaps the reasoner is doing some start-up work.
On 4/10/2021 10:47 AM, Steve Vestal wrote:
> When doing some bench-marking, I ran the same query twice
> back-to-back. The first time took about 4000ms, the second about
> 700ms. The difference remained with optimizations on or off. This
> seems to happen only on the first query, successive pairs of identical
> queries don't show this effect. Is there some sort of start-up
> overhead for the very first query issued?
>
>