Mailing-List: contact users-help@felix.apache.org; run by ezmlm
Precedence: bulk
Reply-To: users@felix.apache.org
Received-SPF: pass (nike.apache.org: domain of pierre.derop@gmail.com
 designates 209.85.215.46 as permitted sender)
MIME-Version: 1.0
In-Reply-To: <etPan.53aad1b1.507ed7ab.119d3@MacBook-Pro-de-Guillaume-2.local>
References: 
 <CAOQaVJw=-u61M2g-SGJ2w2=HUG3JV9CqE+vo4pyb3whvc5CD=A@mail.gmail.com>
	<etPan.53a538ed.70a64e2a.161@Macintosh.local>
	<CAGT3RRN49RkZVtNJHaex1NTEU6OMf2nw28w=KMXNG23XTdi5cg@mail.gmail.com>
	<etPan.53a695a9.2cd89a32.161@Macintosh.local>
	<CAGT3RRPtOCkVVpp7-xsh5-r4FG7xdTi3T+G3uo89XF2rbLjpAg@mail.gmail.com>
	<etPan.53aad1b1.507ed7ab.119d3@MacBook-Pro-de-Guillaume-2.local>
Date: Wed, 25 Jun 2014 16:11:35 +0200
Message-ID: 
 <CAGT3RRNdGcpaPZ8+WC7f3Xio+dHKX5L357wp=+i2tVNdt19w2A@mail.gmail.com>
Subject: Re: IPojo performance benchmark
From: Pierre De Rop <pierre.derop@gmail.com>
To: users@felix.apache.org
Content-Type: multipart/alternative; boundary=001a11c3a08c7288c604fca9a7a6

--001a11c3a08c7288c604fca9a7a6
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

Hi Guillaume;

Thanks for reviewing;

But if I should only use three component types, then when creating each
component instance, is there a way to configure a given component instance
service dependency filter ?

May I provide such filter in the dictionary that I pass to the
ComponentType.createInstance(Dictionary conf) method ? (like passing a
target filter with declarative service ComponentFactory ?). Should I use
another iPOJO api (like an iPojo Factory ?)

In my scenario, I have for example a first "Artist" instance depending on
an "Album" instance with the filter "id=3D1".
then another "Artist" instance depending on another "Album" instance, but
with service filter "id=3D2", etc ...


cheers;
/Pierre


On Wed, Jun 25, 2014 at 3:42 PM, Guillaume Sauthier (OW2) <
guillaume.ow2@gmail.com> wrote:

> Maybe I don't fully understand Java 8 style code, but it looks like you'r=
e
> creating too many PrimitiveComponentType instances, right ?
>
> In iPOJO, a component type is really like a class, you define its contrac=
t
> (the provided specification, aka interface), it's properties (with
> potential default values), =E2=80=A6 And then, you create instances out o=
f 1
> component type (that is a de-facto factory): an instance being a simple
> configuration (a dictionnary).
>
> So, in your benchmark, you should create only 3 PrimitiveComponentType: 1
> for artist, 1 for album and 1 for track.
> Then, you'll have to create instances by providing a configuration with
> instance specific values (like identifier, ...):
>
> http://felix.apache.org/ipojo/api/1.11.0/org/apache/felix/ipojo/api/Compo=
nentType.html#createInstance(java.util.Dictionary)
>
> Thanks
> =E2=80=94G
> Le 25 juin 2014 =C3=A0 12:05:29, Pierre De Rop (pierre.derop@gmail.com) a
> =C3=A9crit:
>
> Hi Clement,
>
> I finally committed a stress test for dependency manager 4 in
>
> https://svn.apache.org/repos/asf/felix/sandbox/pderop/dependencymanager-p=
rototype/
>
> So, I included in it a scenario using iPojo, and it seems that registerin=
g
> around 300 services seems to take an important amount of time, so you mig=
ht
> want to look into the scenario test.
>
> I hope I used the iPojo API correctly, I followed the documentation from
> here:
>
> http://felix.apache.org/documentation/subprojects/apache-felix-ipojo/apac=
he-felix-ipojo-userguide/apache-felix-ipojo-api.html#a-simple-example
> ;
> so if you have time to take a look, let me know if there is something wro=
ng
> with the test and I will correct it.
>
> The stress test is a subproject of dm and is located in
>
> http://svn.apache.org/viewvc/felix/sandbox/pderop/dependencymanager-proto=
type/org.apache.felix.dm.benchmark/
>
> I haved added a README here:
>
> http://svn.apache.org/viewvc/felix/sandbox/pderop/dependencymanager-proto=
type/org.apache.felix.dm.benchmark/README?view=3Dco
>
> And the code used to declare iPojo components (using the iPojo API) is
> located here:
>
>
> http://svn.apache.org/viewvc/felix/sandbox/pderop/dependencymanager-proto=
type/org.apache.felix.dm.benchmark/src/org/apache/felix/dm/benchmark/ipojo/=
IpojoScenario.java?revision=3D1605212&view=3Dco
>
> hope this helps (and also that I used correctly the iPojo APi)
>
> cheers;
> /Pierre
>
>
> On Sun, Jun 22, 2014 at 10:36 AM, Clement Escoffier <
> clement.escoffier@gmail.com> wrote:
>
> > Hi,
> >
> >
> > On 22 juin 2014 at 00:14:26, Pierre De Rop (pierre.derop@gmail.com)
> wrote:
> >
> > Hello Clement,
> >
> > I'm also working on some benchmark related to dependency manager,
> > declarative service, and after reading this post, I tried to incorporat=
e
> a
> > scenario using iPojo in my benchmark.
> > And it seems that I'm also observing a performance problem.
> >
> > I will try to commit somewhere my benchmark program soon in my sandbox,
> > however, I just made a quick jvisualvm, and it looks like I'm spending =
95
> > %
> > of the time in the
> > org.apache.felix.ipojo.IPojoFactory$RetryNameGenerator.generate method.
> >
> >
> > That would make sense. This class is ensuring uniqueness, which is
> > particularly expensive. It's something quite new, and for sure there ar=
e
> > room for optimizations.
> >
> > Clement
> >
> >
> >
> >
> >
> > Here is a stacktrace, if this may help:
> >
> > "Thread-2" #30 prio=3D5 os_prio=3D0 tid=3D0x635f7c00 nid=3D0x47c1 runna=
ble
> > [0x62cfd000]
> > java.lang.Thread.State: RUNNABLE
> > at
> >
> >
> org.apache.felix.ipojo.IPojoFactory$RetryNameGenerator.generate(IPojoFact=
ory.java:1149)
> > at
> >
> >
> org.apache.felix.ipojo.IPojoFactory.createComponentInstance(IPojoFactory.=
java:302)
> > - locked <0x67652198> (a
> > java.util.Collections$SynchronizedRandomAccessList)
> > - locked <0x95055f50> (a org.apache.felix.ipojo.ComponentFactory)
> > at
> >
> >
> org.apache.felix.ipojo.IPojoFactory.createComponentInstance(IPojoFactory.=
java:240)
> > at
> >
> >
> org.apache.felix.ipojo.api.ComponentType.createInstance(ComponentType.jav=
a:79)
> > at
> >
> >
> org.apache.felix.dm.benchmark.ipojo.IPojoActivator$$Lambda$108/3740909.ap=
ply(Unknown
> > Source)
> > at
> > org.apache.felix.dm.benchmark.controller.Helper.lambda$1(Helper.java:85=
)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.Helper$$Lambda$36/6147827.apply(=
Unknown
> > Source)
> > at
> > java.util.stream.ReferencePipeline$3$1.accept(ReferencePipeline.java:19=
3)
> > at
> >
> >
> java.util.ArrayList$ArrayListSpliterator.forEachRemaining(ArrayList.java:=
1359)
> > at
> > java.util.stream.AbstractPipeline.copyInto(AbstractPipeline.java:512)
> > at
> >
> >
> java.util.stream.AbstractPipeline.wrapAndCopyInto(AbstractPipeline.java:5=
02)
> > at
> >
> java.util.stream.ReduceOps$ReduceOp.evaluateSequential(ReduceOps.java:708=
)
> > at
> > java.util.stream.AbstractPipeline.evaluate(AbstractPipeline.java:234)
> > at
> > java.util.stream.ReferencePipeline.collect(ReferencePipeline.java:499)
> > at
> >
> >
> org.apache.felix.dm.benchmark.ipojo.IPojoActivator.__M_start(IPojoActivat=
or.java:96)
> > at
> >
> >
> org.apache.felix.dm.benchmark.ipojo.IPojoActivator.start(IPojoActivator.j=
ava)
> > at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
> > at
> >
> >
> sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java=
:62)
> > at
> >
> >
> sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorI=
mpl.java:43)
> > at java.lang.reflect.Method.invoke(Method.java:483)
> > at org.apache.felix.ipojo.util.Callback.call(Callback.java:237)
> > at org.apache.felix.ipojo.util.Callback.call(Callback.java:193)
> > at
> >
> >
> org.apache.felix.ipojo.handlers.lifecycle.callback.LifecycleCallback.call=
(LifecycleCallback.java:86)
> > at
> >
> >
> org.apache.felix.ipojo.handlers.lifecycle.callback.LifecycleCallbackHandl=
er.__M_stateChanged(LifecycleCallbackHandler.java:162)
> > at
> >
> >
> org.apache.felix.ipojo.handlers.lifecycle.callback.LifecycleCallbackHandl=
er.stateChanged(LifecycleCallbackHandler.java)
> > at
> > org.apache.felix.ipojo.InstanceManager.setState(InstanceManager.java:56=
0)
> > at
> > org.apache.felix.ipojo.InstanceManager.start(InstanceManager.java:440)
> > at
> >
> >
> org.apache.felix.ipojo.ComponentFactory.createInstance(ComponentFactory.j=
ava:179)
> > at
> >
> >
> org.apache.felix.ipojo.IPojoFactory.createComponentInstance(IPojoFactory.=
java:319)
> > - locked <0x6e5340d0> (a org.apache.felix.ipojo.ComponentFactory)
> > at
> >
> >
> org.apache.felix.ipojo.IPojoFactory.createComponentInstance(IPojoFactory.=
java:240)
> > at
> >
> >
> org.apache.felix.ipojo.api.ComponentType.createInstance(ComponentType.jav=
a:79)
> > at
> > org.apache.felix.dm.benchmark.ipojo.Activator.start(Activator.java:31)
> > at
> >
> >
> org.apache.felix.framework.util.SecureAction.startActivator(SecureAction.=
java:645)
> > at org.apache.felix.framework.Felix.activateBundle(Felix.java:2152)
> > at org.apache.felix.framework.Felix.startBundle(Felix.java:2070)
> > at org.apache.felix.framework.BundleImpl.start(BundleImpl.java:976)
> > at org.apache.felix.framework.BundleImpl.start(BundleImpl.java:963)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl.sta=
rtAndStop(BenchmarkControllerImpl.java:190)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl.lam=
bda$17(BenchmarkControllerImpl.java:104)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl$$La=
mbda$17/24713481.run(Unknown
> > Source)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl.dur=
ationOf(BenchmarkControllerImpl.java:229)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl.lam=
bda$11(BenchmarkControllerImpl.java:104)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl$$La=
mbda$15/17116744.applyAsLong(Unknown
> > Source)
> > at java.util.stream.LongPipeline$2$1.accept(LongPipeline.java:214)
> > at
> >
> >
> java.util.stream.Streams$RangeLongSpliterator.forEachRemaining(Streams.ja=
va:235)
> > at
> > java.util.Spliterator$OfLong.forEachRemaining(Spliterator.java:757)
> > at
> > java.util.stream.AbstractPipeline.copyInto(AbstractPipeline.java:512)
> > at
> >
> >
> java.util.stream.AbstractPipeline.wrapAndCopyInto(AbstractPipeline.java:5=
02)
> > at
> > java.util.stream.AbstractPipeline.evaluate(AbstractPipeline.java:576)
> > at
> >
> >
> java.util.stream.AbstractPipeline.evaluateToArrayNode(AbstractPipeline.ja=
va:255)
> > at java.util.stream.LongPipeline.toArray(LongPipeline.java:484)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl.lam=
bda$2(BenchmarkControllerImpl.java:104)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl$$La=
mbda$14/9086622.accept(Unknown
> > Source)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl.lam=
bda$16(BenchmarkControllerImpl.java:169)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl$$La=
mbda$11/14120199.accept(Unknown
> > Source)
> > at java.util.Optional.ifPresent(Optional.java:159)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl.lam=
bda$4(BenchmarkControllerImpl.java:168)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl$$La=
mbda$7/32522449.accept(Unknown
> > Source)
> > at
> >
> >
> java.util.Spliterators$ArraySpliterator.forEachRemaining(Spliterators.jav=
a:948)
> > at
> >
> >
> java.util.stream.ReferencePipeline$Head.forEach(ReferencePipeline.java:58=
0)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl.for=
EachScenarioBundle(BenchmarkControllerImpl.java:166)
> > at
> >
> >
> org.apache.felix.dm.benchmark.controller.impl.BenchmarkControllerImpl.run=
(BenchmarkControllerImpl.java:100)
> > at java.lang.Thread.run(Thread.java:745)
> >
> > I'm using latest released ipojo + java8.
> >
> >
> > (will do my best to commit the benchmark so you can take a look and che=
ck
> > if my test is wrong or if there is really something to be verified
> > somewhere in IPojo).
> >
> >
> >
> > cheers;
> > /Pierre
> >
> >
> > On Sat, Jun 21, 2014 at 9:49 AM, Clement Escoffier <
> > clement.escoffier@gmail.com> wrote:
> >
> > > Hi,
> > >
> > > Thanks,
> > >
> > > I will have a look at it, and come back to you.
> > >
> > > Clement
> > > On 21 juin 2014 at 09:21:46, Xander Uiterlinden (uiterlix@gmail.com)
> > > wrote:
> > >
> > > Hi,
> > >
> > > A colleague of mine just pointed me at this discussion. I'm one of th=
e
> > two
> > > persons that did the presentation on DM performance on the OSGi devco=
n.
> > >
> > > Regarding iPojo, we included it using the documentation on the site b=
ut
> > are
> > > not iPojo experts. I'd be happy to help in figuring out whether the
> > results
> > > are due to the tests or to the iPojo implementation itself.
> > >
> > > The code is at:
> > > https://github.com/sandermak/osgi-dm-shootout/tree/master/benchmark
> > >
> > > Cheers,
> > >
> > > Xander
> > >
> >
>

--001a11c3a08c7288c604fca9a7a6--