db-derby-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Masood Mortazavi <masoodmortaz...@gmail.com>
Subject Re: # of derby instances embedded in a JVM ...
Date Sun, 14 Mar 2010 14:10:46 GMT
Hi -

On Fri, Mar 12, 2010 at 5:15 AM, Knut Anders Hatlen <Knut.Hatlen@sun.com>wrote:

> On 03/12/10 01:05 PM, Masood Mortazavi wrote:
>
>> I am wondering about any limitations regarding the number of Derby
>> instances that can be embedded in a JVM ... (assuming adjustments to
>> heap-size, etc., have been made) ...
>>
>
> Hi Masood,
>
> I haven't tried this myself, but I've heard people talking about having
> about 1000 instances in the same JVM and getting them to run reasonably
> well.
>


I wonder whether all these instances in the same VM can be configured to
have their own distinct storage, and whether that's the configuration used
in the above case.



> The main issues were inefficient use of the heap for caching data (since
> each instance has its own page cache) so that each instance would have to
> run with a very small page cache to prevent exhausting the heap, and a high
> number of background threads (this problem is logged as DERBY-4210).
>


Right.
However, can the background threads issue be avoided if all these instances
have their own distinct "connection" and "storage"?



> Others who have actually tried this might want to add to this or correct me
> if I've got this wrong, though...
>

Yes, it would be good to hear from others who may have tried it.


> Hope this helps,
>
> --
> Knut Anders
>


Yes, it does.
I really appreciate your taking the time to write your note.

Thanks,
-m.

Mime
View raw message