Also, normally publisher mention whether to persist or not messages in
message itself (delivery mode). So based on that MB will process messages
in memory and/or persist to a persistence store. So if it's process in
memory How do we communicate through the MB cluster?

http://activemq.apache.org/what-is-the-difference-between-persistent-and-non-persistent-delivery.html


*Dhanuka Ranasinghe*

Senior Software Engineer
WSO2 Inc. ; http://wso2.com
lean . enterprise . middleware

phone : +94 715381915


On Mon, Jul 21, 2014 at 12:36 PM, Dhanuka Ranasinghe <[email protected]>
wrote:

> IMO, If we gonna keep huge messages as chunks in memory and insert into DB
> as bulk it will heavily affect on MB heap memory. My suggestion is we need
> to handle this case by case. For example, if it's small messages it will be
> efficient to keep in memory while huge messages it will be efficient to
> insert into DB early as possible and let others to use heap memory. For
> this we will have to make this functionality more configurable but again we
> will have to think about how gonna support fail over (probably have to
> change db schema).
>
> *Dhanuka Ranasinghe*
>
> Senior Software Engineer
> WSO2 Inc. ; http://wso2.com
> lean . enterprise . middleware
>
> phone : +94 715381915
>
>
> On Mon, Jul 21, 2014 at 9:00 AM, Asitha Nanayakkara <[email protected]>
> wrote:
>
>> We are planning to insert message chunks as batch insert queries.
>>
>>
>> On Sat, Jul 19, 2014 at 11:00 AM, Dhanuka Ranasinghe <[email protected]>
>> wrote:
>>
>>> Are we going to insert whole message or as chunks
>>> On 18 Jul 2014 18:06, "Asitha Nanayakkara" <[email protected]> wrote:
>>>
>>>>  Hi,
>>>>
>>>> Following is the RDBMS design for WSO2 MB 3.0.0
>>>>
>>>> Messages model
>>>>
>>>> ​Message metadata model
>>>>
>>>>
>>>> ​
>>>> Following are the concerns came across in the discussion
>>>>
>>>> *- Why we use reference counting for message meta data?*
>>>>
>>>> Reference counting is needed to delete topic messages from the database
>>>> reliably in a cluster deployment
>>>>
>>>> *-​ How to manage a large tables like Messages table?*
>>>>
>>>> for Messages table use database partitioning
>>>>
>>>> For Metadata queries there will be no SQL joins, hence even if the
>>>> table would grow large that won't be an issue.
>>>>
>>>> Inserts and delete operation can be done as batch operations.
>>>>
>>>> *- Following option to save metadata was rejected due to following
>>>> reasons*
>>>>
>>>>
>>>>
>>>> In the above design MB instance will create Node_Queue tables and
>>>> Topic_Node_Queue
>>>> tables when each node connects to a cluster. This design was rejected
>>>> due to following reasons.
>>>> It's DB admins tasks to create and delete tables. MB should not modify
>>>> schema when joining to the cluster. There will be instances where MB users
>>>> might not have privileges to create tables.
>>>>
>>>> *- Supporting several SQL implementations.*
>>>>
>>>> Since we are using simple SQL operations those will not become an
>>>> issue.
>>>>
>>>>
>>>>
>>>>
>>>> Thanks
>>>> --
>>>> *Asitha Nanayakkara*
>>>> Software Engineer
>>>> WSO2, Inc. http://wso2.com/
>>>> Mob: + 94 77 85 30 682
>>>> ​​
>>>>
>>>> _______________________________________________
>>>> Architecture mailing list
>>>> [email protected]
>>>> https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture
>>>>
>>>>
>>> _______________________________________________
>>> Architecture mailing list
>>> [email protected]
>>> https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture
>>>
>>>
>>
>>
>> --
>> *Asitha Nanayakkara*
>>  Software Engineer
>> WSO2, Inc. http://wso2.com/
>> Mob: + 94 77 85 30 682
>>
>>
>> _______________________________________________
>> Architecture mailing list
>> [email protected]
>> https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture
>>
>>
>
_______________________________________________
Architecture mailing list
[email protected]
https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture

Reply via email to