openjpa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marc Prud'hommeaux <mprud...@apache.org>
Subject Re: sql statement batching
Date Thu, 18 Jan 2007 02:00:21 GMT
Kevin-

On Jan 15, 2007, at 6:39 AM, Kevin Sutter wrote:

> Marc,
> Concerning the infrastructure for sql batching support...  Do you  
> mean that
> this support is pluggable into the OpenJPA runtime?  That is, is  
> there a
> pluggable interface for other vendors to provide support for sql  
> batching?
> Or, is it just that the runtime could be updated to support sql  
> batching
> (and then the source would be out of synch with the OpenJPA  
> repository)?

The former. Kodo is built on top of OpenJPA with no source  
modification or patches (using only the ProductDerivation framework).  
Without going into too many details about the actual implementation,  
it is mostly done using subclasses of the PreparedStatementManager  
and OperationOrderUpdateManager.


> I'm just trying to assess the impact for this "removed" function.   
> If each
> vendor that incorporates OpenJPA needs to modify their version of the
> runtime to provide sql batching support (and it's a common  
> feature), then
> the consistency of OpenJPA will be impacted.

I sympathize with that concern. All I can really say is that the  
holdbacks to the OpenJPA contributions were carefully considered to  
maximize functionality rather than performance.


> From Dave's note, it seems like we have a mistake in the user's  
> guide (at a
> minimum).  But, even if the "addBatch limit" is taken out of the  
> user's
> guide, then as each vendor adds this support back in (for performance
> reasons), the configuration of this feature may also vary from one  
> vendor to
> the next.

It is indeed a mistake in the documentation. It should be removed.  
The exact configuration property may vary from vendor to vendor.


> I'm still trying to gather some "real data" to demonstrate the  
> performance
> advantages of sql batching (or lack there of).  But, if Dave's  
> comments are
> accurate, then maybe we need to consider re-introducing the sql  
> batching
> support.  Not only for performance reasons, but also for runtime and
> configuration consistency reasons.

Clearly it would be simplest and most beneficial if the code for  
batching was added to the contribution. I can't really comment on the  
likelihood of whether that will ever happen or not, but I agree that  
it would be nice.


> My two cents worth...  :-)
>
> Kevin
>
> On 1/11/07, Marc Prud'hommeaux <mprudhom@apache.org> wrote:
>>
>> David-
>>
>> Batching functionality was not included in the OpenJPA code
>> contribution made by BEA, but you are correct that the infrastructure
>> necessary for its support is in place.
>>
>> I can't comment on whether BEA could be persuaded to contribute that
>> code as well beyond saying that they are always open to community
>> suggestions and feedback.
>>
>>
>>
>> On Jan 10, 2007, at 5:11 PM, David Wisneski wrote:
>>
>> > Does OpenJPA support sql batching (i.e. the jdbc addBatch api on a
>> > prepared Statement)?  The user's  guide shows how to set the  
>> addBatch
>> > limit, but  it does not work.  It would appear that if there was  
>> code
>> > to do this it was removed.
>> >
>> > I think that most ORM runtimes support this feature and we will  
>> be at
>> > a competitive disadvantage with Hibernate and TopLink without  
>> it.  It
>> > is very important when creating and persist many new object of the
>> > same type in one transaction (i.e. persisting 1000 new Customer  
>> beans)
>> > or doing many updates ( updating the salary on 1000 Employee  
>> beans).
>> > I know that addBatch performance vs. doing multiple jdbc  
>> executeUpdate
>> > makes a big differrent in DB2 -- and I would assume also other
>> > databases.
>> >
>> > If the code was there and removed -- any thoughts about adding it
>> > back in?  Otherwise we would most likely have to develop it for our
>> > enterprise customer.
>>
>>


Mime
View raw message