Hi Sylvain,
What I can not run such many tests, what is RAM amount you are interesting
for.
Lets say 2 GB will good enough ?
Actually I can push benchmark on github so you will be able to run it if
you do not mind.


On Fri, Jan 31, 2014 at 12:26 PM, Sylvain Spinelli <
[email protected]> wrote:

>  Hi Andrey,
>
> interesting benchmark.
>
> The point interesting for me is memory consumption.
> With plocal how memory is configured ? Could you execute the same test
> with max memory settings divided by 2, by 4, by 8... ? In order to estimate
> when times became crazy or OOM is thrown.
>
> Why this point is important for me ? Of course in dedicated server with 64
> or 128Gb RAM it's not so important, but I have 2 other use cases :
>
>    - standalone application (with orient embedded running on the client
>    computer). Memory settings must be automatic/dynamic and should start to
>    work with very few memory. Of course if the user create millions records,
>    it is normal that the application requires more memory.
>     - low-cost cloud. We plan to deploy a new very low-cost solution with
>    shared server. Each user has his own webapp and OriendDb instance but
>    running in the same Tomcat server. So we must fix memory settings, with
>    quite low values for each user.
>
> That's why evaluate how performance degrades with less memory is very
> interesting.
>
> Thanks,
> Sylvain
>
>
> Le 31/01/2014 10:25, Andrey Lomakin a écrit :
>
> Couple of words about benchmark.
>
>  We imported wikipedia data using OrientGraph database, to be sure that
> we have no duplications unique index by page key was created.
>
>  First benchmark that is benchmark which uses 1.6.4 version of OrientDB
> and second benchmark uses version from
> https://github.com/orientechnologies/orientdb/tree/rid-set-sbtree branch.
>
>
>
>
> On Fri, Jan 31, 2014 at 11:22 AM, Andrey Lomakin <[email protected]
> > wrote:
>
>> Hi,
>> Here is difference in implementations in tx mode, local + mvrbtree +
>> unique index (old) and plocal + ridbag + unique index (new).
>>
>>
>>
>>  On Fri, Jan 31, 2014 at 10:37 AM, Andrey Lomakin <
>> [email protected]> wrote:
>>
>>> Hi,
>>> Connection pool can be used from document database.
>>> Create document database and pass it in OrientGraph it is a bit
>>> inconvenient but we did benchmark of import of wikipedia data structure it
>>> is 10-15 times faster on some data sets.
>>>
>>>
>>> On Thu, Jan 30, 2014 at 10:59 PM, Riccardo Tasso <
>>> [email protected]> wrote:
>>>
>>>> And what about the ConnectionPool? If I remember well it was usable
>>>> only from OGraphDatabase.
>>>>
>>>>  Thanks,
>>>>    Riccardo
>>>>
>>>>
>>>>  2014-01-30 Claudio Tesoriero <[email protected]>:
>>>>
>>>>  Thanks!
>>>>>  Il 30/gen/2014 17:33 "Andrey Lomakin" <[email protected]> ha
>>>>> scritto:
>>>>>
>>>>>>
>>>>>>  Hi,
>>>>>> To create graph db in such way just use
>>>>>>
>>>>>>  db = Orient.instance().getDatabaseFactory().createDatabase("graph", 
>>>>>> "plocal:"
>>>>>> + dbPath)
>>>>>>  if (!db.exists()) {
>>>>>>  info("DB does not exist, BaasBox will create a new one");
>>>>>>  db.create();
>>>>>> }
>>>>>>
>>>>>>  About the second question, you have several variants.
>>>>>>
>>>>>>  Two of them:
>>>>>>  1.   db.addVertex("class:" + CLASS_VERTEX_NAME);
>>>>>>  2.  db.addVertex(CLASS_VERTEX_NAME, null);
>>>>>>
>>>>>>
>>>>>>
>>>>>> On Thu, Jan 30, 2014 at 5:30 PM, Claudio <[email protected]>wrote:
>>>>>>
>>>>>>> Hi Luca,
>>>>>>> we use the old OGraphDatabase API  to create a new fresh graph
>>>>>>> database if it does not exists when our application starts.
>>>>>>>  db = new OGraphDatabase( "plocal:" + dbPath ) ;
>>>>>>> if (!db.exists()) {
>>>>>>>  info("DB does not exist, BaasBox will create a new one");
>>>>>>>  db.create();
>>>>>>> }
>>>>>>>
>>>>>>>  As far I know the OrientGraph API hasn't got the exists() method,
>>>>>>> on the contrary, its constructor raises an error if the specified DB  
>>>>>>> does
>>>>>>> not exists.
>>>>>>> Any advice on this?
>>>>>>>
>>>>>>>  Furthermore we use it to create vertices:
>>>>>>> ODocument vertex = db.createVertex(CLASS_VERTEX_NAME);
>>>>>>>
>>>>>>> Inspecting the OrientGraph API, I cannot find a similar API. I just
>>>>>>> found the addVertex(iClassName, iClusterName) that accepts two
>>>>>>> parameters instead of just the class name
>>>>>>>
>>>>>>>  Thanks,
>>>>>>> Claudio
>>>>>>>
>>>>>>>
>>>>>>> Il giorno mercoledì 29 gennaio 2014 11:26:21 UTC+1, Lvc@ ha
>>>>>>> scritto:
>>>>>>>
>>>>>>>> Hi all,
>>>>>>>> we're very close to merge the new storage structure to the 1.7
>>>>>>>> branch. Andrey has some *awesome numbers *about the performance of
>>>>>>>> the new engine in terms of benchmarks, he will share them with us on 
>>>>>>>> this
>>>>>>>> group very soon.
>>>>>>>>
>>>>>>>>  But in order to complete the merge, we should *remove the old
>>>>>>>> Graph API: OGraphDatabase*! This is mainly for 2 big reasons:
>>>>>>>>
>>>>>>>>    1. it's not compatible with *Lightweight* edges and other
>>>>>>>>    improvements we introduced since version 1.4.x
>>>>>>>>     2. it heavily uses *MVRB-Tree indexes* so it wouldn't scale on
>>>>>>>>    the new storage structure
>>>>>>>>
>>>>>>>>  So after a while we deprecated the old Graph API, it's time to
>>>>>>>> removed it from the v1.7 code base. I know some users are still sticked
>>>>>>>> with this API, they can decide to:
>>>>>>>>
>>>>>>>>    1. Migrate to the Blueprints API (well documented on OrientDB &
>>>>>>>>    TinkerPop sites)
>>>>>>>>    2. Stay with old API and remain sticked to last OrientDB v1.6.x.
>>>>>>>>
>>>>>>>>  Obviously we strongly suggest 1st solution!
>>>>>>>>
>>>>>>>>  For the users want to migrate we will provide a new guide to
>>>>>>>> migrate the old code (very easy, most of the concepts are the same). In
>>>>>>>> this way they could benefits of HUGE improvement with last version of
>>>>>>>> OrientDB, specially with big databases.
>>>>>>>>
>>>>>>>>  Any comments are welcome :-)
>>>>>>>>
>>>>>>>>  Lvc@
>>>>>>>>
>>>>>>>>      --
>>>>>>>
>>>>>>> ---
>>>>>>> You received this message because you are subscribed to the Google
>>>>>>> Groups "OrientDB" group.
>>>>>>> To unsubscribe from this group and stop receiving emails from it,
>>>>>>> send an email to [email protected].
>>>>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>  --
>>>>>> Best regards,
>>>>>> Andrey Lomakin.
>>>>>>
>>>>>> Orient Technologies
>>>>>> the Company behind OrientDB
>>>>>>
>>>>>>   --
>>>>>>
>>>>>> ---
>>>>>>  You received this message because you are subscribed to a topic in
>>>>>> the Google Groups "OrientDB" group.
>>>>>> To unsubscribe from this topic, visit
>>>>>> https://groups.google.com/d/topic/orient-database/oFYb96sM7Jg/unsubscribe
>>>>>> .
>>>>>> To unsubscribe from this group and all its topics, send an email to
>>>>>> [email protected].
>>>>>>
>>>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>>>
>>>>>   --
>>>>>
>>>>> ---
>>>>> You received this message because you are subscribed to the Google
>>>>> Groups "OrientDB" group.
>>>>> To unsubscribe from this group and stop receiving emails from it, send
>>>>> an email to [email protected].
>>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>>
>>>>
>>>>   --
>>>>
>>>> ---
>>>> You received this message because you are subscribed to the Google
>>>> Groups "OrientDB" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send
>>>> an email to [email protected].
>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>
>>>
>>>
>>>
>>>  --
>>> Best regards,
>>> Andrey Lomakin.
>>>
>>> Orient Technologies
>>> the Company behind OrientDB
>>>
>>>
>>
>>
>>  --
>> Best regards,
>> Andrey Lomakin.
>>
>> Orient Technologies
>> the Company behind OrientDB
>>
>>
>
>
>  --
> Best regards,
> Andrey Lomakin.
>
> Orient Technologies
> the Company behind OrientDB
>
>  --
>
> ---
> You received this message because you are subscribed to the Google Groups
> "OrientDB" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> For more options, visit https://groups.google.com/groups/opt_out.
>
>
>  --
>
> ---
> You received this message because you are subscribed to the Google Groups
> "OrientDB" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> For more options, visit https://groups.google.com/groups/opt_out.
>



-- 
Best regards,
Andrey Lomakin.

Orient Technologies
the Company behind OrientDB

-- 

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

Reply via email to