this was done with 2 node MB cluster. We need to test and decide with
larger cluster sizes.


On Thu, Apr 30, 2015 at 5:59 PM, Afkham Azeez <[email protected]> wrote:

> Was this test done with a single MB instance or while having more than 1
> MB instance in the same cluster?
>
> On Thu, Apr 30, 2015 at 5:56 PM, Ramith Jayasinghe <[email protected]>
> wrote:
>
>> HI Shanker,
>>  We keep lots of state in Hazel-cast IMaps for MB.
>>  What we want to achieve by this experiment is to figure out weather we
>> can reduce the inconsistencies that could happen during failover scenarios
>> (we are yet to do a failover  test round with this, will keep everyone
>> updated with the progress).
>>
>>  May be we can make this the default way of keeping slot based
>> information if things work better compared hazelcast approach ( - again we
>> are going to determine by testing the performance and failover scenarios).
>>
>>  However, I prefer to keep both options available for the users and then
>> eventually deprecate one approach over time.
>>
>>
>> On Thu, Apr 30, 2015 at 5:28 PM, Selvaratnam Uthaiyashankar <
>> [email protected]> wrote:
>>
>>>
>>>
>>> On Thu, Apr 30, 2015 at 5:00 PM, Anuja Herath <[email protected]> wrote:
>>>
>>>> Hi,
>>>>
>>>> Is it single node MB?
>>>>
>>>> No, It is a cluster. I used two MB nodes for testing. Publisher in one
>>>> node and Subscriber in another node.
>>>>
>>>> Database is in the same machine?
>>>>
>>>> For MS SQL, I used a separate machine with database server.
>>>> For MySQL, I used database in the same machine.
>>>>
>>>> Any reason why we used MSSQL?
>>>>
>>>> MB performs faster in MS SQL, So initially, we wanted to test new
>>>> implementation in that best case.
>>>> We are planning to continue testing with other databases as well.
>>>>
>>>> Result looks there are not much difference between the two. So, what is
>>>> the conclusion?
>>>>
>>>> MB is heavily depend on Hazelcast and use several hazelcast maps. And
>>>> also we faced several problems related to that.
>>>> This implementation was done as an experiment to release that heavy
>>>> dependency on Hazelcast.
>>>> We did not expect to increase performance with this implementation. But
>>>> at the same time, we did not want to decrease performance with a new
>>>> implementation.
>>>> So with these preliminary test results, we concluded that there is no
>>>> much different in performance between using RDBMS and Hazelcast for storing
>>>> slot information.
>>>>
>>>> If the new implementation works with further testing, we will add an
>>>> option to select either RDBMS or Hazelcast for storing slot information.
>>>>
>>>
>>>
>>> So, this is going to be one more configuration? How the users will know
>>> which one is good for them?
>>>
>>>
>>>
>>>>
>>>> Thanks,
>>>> Anuja
>>>>
>>>> On Thu, Apr 30, 2015 at 4:29 PM, Selvaratnam Uthaiyashankar <
>>>> [email protected]> wrote:
>>>>
>>>>> Is it single node MB?
>>>>> Database is in the same machine?
>>>>> Any reason why we used MSSQL?
>>>>> Result looks there are not much difference between the two. So, what
>>>>> is the conclusion?
>>>>>
>>>>> On Thu, Apr 30, 2015 at 12:12 PM, Anuja Herath <[email protected]> wrote:
>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> I have implemented storing slot information using RDBMS. Previously
>>>>>> MB used Hazelcast for storing these slot information. I have attached
>>>>>> performance comparison related to above implementation.
>>>>>>
>>>>>> [1] Performance Results -
>>>>>> https://docs.google.com/a/wso2.com/spreadsheets/d/1O9Mr-TiZaIyOBox4KhxQkcy_puG_i9JXe4xbPY3QPg4/edit?usp=sharing
>>>>>>
>>>>>> [2] Redmine Project - https://redmine.wso2.com/issues/3801
>>>>>>
>>>>>> Thanks,
>>>>>> Anuja
>>>>>>
>>>>>> --
>>>>>> Anuja Herath
>>>>>> *Software Engineer*
>>>>>> *WSO2, Inc.*
>>>>>> Mobile : +94 (0)71 429 8861
>>>>>>
>>>>>> _______________________________________________
>>>>>> Architecture mailing list
>>>>>> [email protected]
>>>>>> https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> S.Uthaiyashankar
>>>>> VP Engineering
>>>>> WSO2 Inc.
>>>>> http://wso2.com/ - "lean . enterprise . middleware"
>>>>>
>>>>> Phone: +94 714897591
>>>>>
>>>>> _______________________________________________
>>>>> Architecture mailing list
>>>>> [email protected]
>>>>> https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> Anuja Herath
>>>> *Software Engineer*
>>>> *WSO2, Inc.*
>>>> Mobile : +94 (0)71 429 8861
>>>>
>>>
>>>
>>>
>>> --
>>> S.Uthaiyashankar
>>> VP Engineering
>>> WSO2 Inc.
>>> http://wso2.com/ - "lean . enterprise . middleware"
>>>
>>> Phone: +94 714897591
>>>
>>> _______________________________________________
>>> Architecture mailing list
>>> [email protected]
>>> https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture
>>>
>>>
>>
>>
>> --
>> Ramith Jayasinghe
>> Technical Lead
>> WSO2 Inc., http://wso2.com
>> lean.enterprise.middleware
>>
>> E: [email protected]
>> P: +94 777542851
>>
>>
>> _______________________________________________
>> Architecture mailing list
>> [email protected]
>> https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture
>>
>>
>
>
> --
> *Afkham Azeez*
> Director of Architecture; WSO2, Inc.; http://wso2.com
> Member; Apache Software Foundation; http://www.apache.org/
> * <http://www.apache.org/>*
> *email: **[email protected]* <[email protected]>
> * cell: +94 77 3320919 <%2B94%2077%203320919>blog: *
> *http://blog.afkham.org* <http://blog.afkham.org>
> *twitter: **http://twitter.com/afkham_azeez*
> <http://twitter.com/afkham_azeez>
> *linked-in: **http://lk.linkedin.com/in/afkhamazeez
> <http://lk.linkedin.com/in/afkhamazeez>*
>
> *Lean . Enterprise . Middleware*
>
> _______________________________________________
> Architecture mailing list
> [email protected]
> https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture
>
>


-- 
Ramith Jayasinghe
Technical Lead
WSO2 Inc., http://wso2.com
lean.enterprise.middleware

E: [email protected]
P: +94 777542851
_______________________________________________
Architecture mailing list
[email protected]
https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture

Reply via email to