Hello!

In streaming mode, INSERTs should not block, so you may reuse the
connection. Still, it would not hurt to benchmark both approaches, if you
can.

Regards,
-- 
Ilya Kasnacheev


сб, 2 мая 2020 г. в 01:42, narges saleh <[email protected]>:

> Hi All,
>  If I use client connection configuration to set the number of threads for
> a JDBC connection,  and use the connection with multiple insert statements
> (with streaming set to true), to multiple different caches, are the inserts
> stacked because the connection is shared? Should I create multiple JDBC
> connections and switch them across the statements?
> thanks.
>
> On Thu, Apr 16, 2020 at 12:33 PM Evgenii Zhuravlev <
> [email protected]> wrote:
>
>> As I said, if you use only DataStreamer, without jdbc, just a plain,
>> key-value IgniteDataStreamer, then, you should have only one instance per
>> cache. It will give you the better performance. This one streamer can be
>> used from multiple threads.
>>
>> чт, 16 апр. 2020 г. в 09:54, narges saleh <[email protected]>:
>>
>>> I am sorry for mixing these two up.
>>> I am asking if I were to use binaryobject builder with datastreamer in
>>> place on jdbc connection would/should I create a pool of the streamer
>>> objects. From your answers, the answer seems to be yes. thank you.
>>>
>>> On Thu, Apr 16, 2020 at 9:07 AM Evgenii Zhuravlev <
>>> [email protected]> wrote:
>>>
>>>> You said that you use Binary Object Builder, so, I thought that you use
>>>> key value API and data streamers. I don't really understand now you use
>>>> BinaryObjectBuilder with thin JDBC client.
>>>>
>>>> >What if I have a persistent connection that sends data continuously?
>>>> Should I hold on the instance of the streamer (for a particular cache), or
>>>> recreate a new one once a new load of data arrives?
>>>> If you're loading data continuously, it makes sense to store data
>>>> streamer instance somewhere and just reuse it, avoiding recreating it each
>>>> time. I
>>>>
>>>> >Are you saying have the data streamed to the streamer via multiple
>>>> connections, across multiple threads?
>>>> If you use just a simple IgniteDataStreamer, you can use it from
>>>> multiple threads(use addData from multiple threads) to increase the
>>>> throughput.
>>>> Evgenii
>>>>
>>>> ср, 15 апр. 2020 г. в 12:07, narges saleh <[email protected]>:
>>>>
>>>>> Hello Evgenii,
>>>>>
>>>>> I am not sure what you mean by reuse a data streamer from multiple
>>>>> threads.  I have data being constantly "streamed" to the streamer via a
>>>>> connection. Are you saying have the data streamed to the streamer via
>>>>> multiple connections, across multiple threads?
>>>>> What if I have a persistent connection that sends data continuously?
>>>>> Should I hold on the instance of the streamer (for a particular cache), or
>>>>> recreate a new one once a new load of data arrives?
>>>>>
>>>>> On Wed, Apr 15, 2020 at 1:17 PM Evgenii Zhuravlev <
>>>>> [email protected]> wrote:
>>>>>
>>>>>> > Should I create a pool of data streamers (a few for each cache)?
>>>>>> If you use just KV API, it's better to have only one data streamer
>>>>>> per cache and reuse it from multiple threads - it will give you the best
>>>>>> performance.
>>>>>>
>>>>>> Evgenii
>>>>>>
>>>>>> ср, 15 апр. 2020 г. в 04:53, narges saleh <[email protected]>:
>>>>>>
>>>>>>> Please note that in my case, the streamers are running on the server
>>>>>>> side (as part of different services).
>>>>>>>
>>>>>>> On Wed, Apr 15, 2020 at 6:46 AM narges saleh <[email protected]>
>>>>>>> wrote:
>>>>>>>
>>>>>>>> So, in effect, I'll be having a pool of streamers, right?
>>>>>>>> Would this still be the case if I am using BinaryObjectBuilder to
>>>>>>>> build objects to stream the data to a few caches? Should I create a 
>>>>>>>> pool of
>>>>>>>> data streamers (a few for each cache)?
>>>>>>>> I don't want to have to create a new object builder and data
>>>>>>>> streamer if I am inserting to the same cache over and over.
>>>>>>>>
>>>>>>>> On Tue, Apr 14, 2020 at 11:56 AM Evgenii Zhuravlev <
>>>>>>>> [email protected]> wrote:
>>>>>>>>
>>>>>>>>> For each connection, on node side will be created its own
>>>>>>>>> datastreamer. I think it makes sense to try pooling for data load, 
>>>>>>>>> but you
>>>>>>>>> will need to measure everything, since the pool size depends on the 
>>>>>>>>> lot of
>>>>>>>>> things
>>>>>>>>>
>>>>>>>>> вт, 14 апр. 2020 г. в 07:31, narges saleh <[email protected]>:
>>>>>>>>>
>>>>>>>>>> Yes, Evgenii.
>>>>>>>>>>
>>>>>>>>>> On Mon, Apr 13, 2020 at 10:06 PM Evgenii Zhuravlev <
>>>>>>>>>> [email protected]> wrote:
>>>>>>>>>>
>>>>>>>>>>> Hi,
>>>>>>>>>>>
>>>>>>>>>>> Do you use STREAMING MODE for thin JDBC driver?
>>>>>>>>>>>
>>>>>>>>>>> Evgenii
>>>>>>>>>>>
>>>>>>>>>>> пн, 13 апр. 2020 г. в 19:33, narges saleh <[email protected]
>>>>>>>>>>> >:
>>>>>>>>>>>
>>>>>>>>>>>> Thanks Alex. I will study the links you provided.
>>>>>>>>>>>>
>>>>>>>>>>>> I read somewhere that jdbc datasource is available via Ignite
>>>>>>>>>>>> JDBC, (which should provide connection pooling).
>>>>>>>>>>>>
>>>>>>>>>>>> On Mon, Apr 13, 2020 at 12:31 PM akorensh <
>>>>>>>>>>>> [email protected]> wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>>   At this point you need to implement connection pooling
>>>>>>>>>>>>> yourself.
>>>>>>>>>>>>>   Use
>>>>>>>>>>>>>
>>>>>>>>>>>>> https://ignite.apache.org/releases/latest/javadoc/org/apache/ignite/configuration/ClientConnectorConfiguration.html#setThreadPoolSize-int-
>>>>>>>>>>>>>   to specify number of threads Ignite creates to service
>>>>>>>>>>>>> connection
>>>>>>>>>>>>> requests.
>>>>>>>>>>>>>
>>>>>>>>>>>>>   Each new connection will be handled by a separate thread
>>>>>>>>>>>>> inside
>>>>>>>>>>>>> Ignite(maxed out a threadPoolSize - as described above)
>>>>>>>>>>>>>
>>>>>>>>>>>>>   ClientConnectorConfiguration is set inside
>>>>>>>>>>>>> IgniteConfiguration:
>>>>>>>>>>>>>
>>>>>>>>>>>>> https://ignite.apache.org/releases/latest/javadoc/org/apache/ignite/configuration/IgniteConfiguration.html#setClientConnectorConfiguration-org.apache.ignite.configuration.ClientConnectorConfiguration-
>>>>>>>>>>>>>
>>>>>>>>>>>>>   More info:
>>>>>>>>>>>>>
>>>>>>>>>>>>> https://www.gridgain.com/docs/latest/developers-guide/SQL/JDBC/jdbc-driver#cluster-configuration
>>>>>>>>>>>>>
>>>>>>>>>>>>> Thanks, Alex
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> --
>>>>>>>>>>>>> Sent from: http://apache-ignite-users.70518.x6.nabble.com/
>>>>>>>>>>>>>
>>>>>>>>>>>>

Reply via email to